Download - research.vu.nl dissertation.pdfVRIJE UNIVERSITEIT Constructions, constraints, and construal Adpositions in Dutch ACADEMISCH PROEFSCHRIFT ter verkrijging van de graad Doctor aan de

Constructions, constraints, and construal

Adpositions in Dutch

Published by LOT phone: +31 30 253 6006 Janskerkhof 13 fax: +31 30 253 6406 3512 BL Utrecht e-mail: [email protected] The Netherlands http://www.lotschool.nl Cover illustration: Watanabe Seitei, c. 1920, ‘Two frogs in a pond with reeds’ (detail). Photography: C.P.J. van der Peet bv Japanese Prints. ISBN 978-90-78328-68-1 NUR 616 Copyright © 2008: Maaike Beliën. All rights reserved.

VRIJE UNIVERSITEIT

Constructions, constraints, and construal

Adpositions in Dutch

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam, op gezag van de rector magnificus

prof.dr. L.M. Bouter, in het openbaar te verdedigen

ten overstaan van de promotiecommissie van de faculteit der Letteren

op donderdag 4 december 2008 om 13.45 uur in het auditorium van de universiteit,

De Boelelaan 1105

door

Maaike Laura Beliën

geboren te Heemskerk

promotor: prof.dr. Th.A.J.M. Janssen copromotor: dr. F.C. van der Leek

Acknowledgments This has been a most enjoyable project. There are quite a few people who have contributed to it in some way. I would like to take the opportunity to thank them here.

First of all, I would like to thank Theo Janssen and Frederike van der Leek, perhaps most of all for making me feel that they have enjoyed this project as much as I have. Their comments, suggestions, and constant support have been invaluable. I am grateful to Theo for his relentless optimism, his sharp observations about language data, and his ability to reduce seemingly insurmountable problems to manageable proportions. I am grateful to Frederike for her inspirational classes on cognitive linguistics, for all our discussions over the years, in which she was always three steps ahead of me, and for her meticulously critical eye.

I would also like to thank the members of my reading committee: Melissa Bowerman, Hubert Cuyckens, Marjon Helmantel, Lachlan Mackenzie, and Arie Verhagen. I am grateful to them for taking the time to read my manuscript.

The idea for this dissertation arose during the academic year 1994-1995, when I was an exchange student at UC Berkeley. I am grateful to the faculty and students of the UC Berkeley Linguistics Department for their formative influence on my linguistic thinking, particularly to those whose classes I took: Charles Fillmore, Paul Kay, James Matisoff, Eve Sweetser, and Dan Slobin. I would especially like to thank Eve Sweetser for making me feel very welcome in this bustling intellectual community.

The bulk of the research presented in this dissertation, carried out during my years at the Vrije Universiteit Amsterdam, was funded by the Netherlands Organization for Scientific Research (NWO, project 200-50-083). I feel very fortunate that I was able to continue the project as a member of the Amsterdam Center for Language and Communication (ACLC) at the University of Amsterdam, and that I am finishing it now at the Institute for Technology and Communication of the Technical University Delft.

When I started my project at the Vrije Universiteit, I expected it to be me, books, and a computer. Little did I know. I have fond memories of the daily Ph.D./postdoc lunches in the delightful company of Hilde van Aken, Frank Beths, Aleth Bolt, Corrien Blom, Ronny Boogaart, Mirjam Ernestus, Lidewij van Gils, Marjan Huisman, Bettelou Los, Margit Rem and Ariane Zwiers. I also very much enjoyed sharing an office with Ronny Boogaart and Marjan Huisman as well as chatting over coffee with Gert Bos.

In 2001 I spent three months as a visiting scholar at the Linguistics Department of UC San Diego, for which I am indebted to Ronald Langacker. I would like to thank him and Gilles Fauconnier for their stimulating classes and their willingness to discuss my work with them. I would also like to thank Tore Nesset, Esther Pascual, Anne Sumnicht, and Inge Zwitserlood for their enjoyable company.

I also wish to express my gratitude towards my colleagues at the English Department of the University of Amsterdam. I would particularly like to thank Evelien Keizer, who I could always turn to with questions on linguistic, teaching, or other matters. I thank Olga Fischer for her faith in me. Thanks also to the core of the always inspiring language acquisition team: Anne Bannink, Manon van der Laaken, Rosalie Mesker, and Roos van der Zwaard. Special thanks, finally, to Jetty Peterse and Henny de Boer for their warm-hearted contribution to a very pleasant working environment.

For their interest in my project over the years, I thank Ad Foolen, Wim Honselaar, Marco Last, Enrique Palancar, Esther Pascual, Co Vet, and Sarah van Vliet. I am grateful also to the late Paul Werth for his classes on discourse with their emphasis on authentic language data. I further would like to thank Eric Akkerman for helping me find a suitable language corpus and for showing me how to use it. Thanks also to my new colleagues in Delft for giving me a warm welcome. I am especially grateful to Merel Keijzer and Liza Berry for showing me the ropes and answering all my questions.

I am grateful to my dear friends Marieke Timmerman and Anna Bekius for agreeing to be my paranimfen. I know that Nelleke Vercouteren would have loved to see this, and I miss her. I thank my parents for their loving support. I would especially like to thank my mother for always being willing to look after Joppe and Wieke when I felt I needed more time to work, and I thank my father for making academia look like a fun line of work. Finally, I want to thank Joppe and Wieke for forcing me to be more efficient with my time, and Bart for giving me extra time to attend conferences or write. Most of all, I thank the three of them for our lovely life together.

Contents 1 Introduction ..................................................................................... 11 1.1 A long-standing problem in Dutch linguistics .......................... 11 1.2 A different approach.................................................................. 13 1.3 Overview of the study ............................................................... 15 2 Postposition or particle: Constituency tests...........................17 2.1 Introduction ............................................................................... 17 2.2 No verb to form an SCV with ................................................... 18 2.3 Non-subject nominal and adposition not adjacent .................... 18 2.4 Topicalization............................................................................ 20 2.5 Pronominalization ..................................................................... 21 2.6 Passivization.............................................................................. 23 2.7 Auxiliary choice ........................................................................ 24 2.8 Nominalization.......................................................................... 25 2.9 Caused motion........................................................................... 25 2.10 Coordination.............................................................................. 26 2.11 Conclusion................................................................................. 27 3 Previous semantic studies............................................................. 29 3.1 Introduction ............................................................................... 29 3.2 Perfective and imperfective....................................................... 29 3.3 Location and direction............................................................... 31 3.4 Four problems ........................................................................... 36 3.5 Conclusion................................................................................. 45 4 Adpositional semantics: The case of over.................................. 47 4.1 Introduction ............................................................................... 47 4.2 A lexical network analysis of Dutch over................................. 48 4.3 Dutch over: a single constraint.................................................. 54 4.3.1 Introduction ..................................................................... 54 4.3.2 Linguistic meanings as constraints.................................. 55 4.3.3 Dimensions of trajector and landmark............................. 56 4.3.4 Contact............................................................................. 57 4.3.5 Boundary crossing........................................................... 58 4.3.6 Static or moving trajector ................................................ 61 4.3.7 Covering .......................................................................... 66 4.3.8 A force............................................................................. 66

4.4 Discussion ................................................................................. 69 4.4.1 Methodology ................................................................... 69 4.4.2 Experimental evidence .................................................... 70 4.4.3 Context-independent meanings ....................................... 71 4.4.4 Conventionality ............................................................... 72 4.5 Conclusion................................................................................. 73 5 Prepositional phrases and auxiliary choice.............................. 75 5.1 Introduction ............................................................................... 75 5.2 Auxiliary choice and PPs .......................................................... 75 5.3 Problematic attested data........................................................... 79 5.4 PreCs: a cognitive-grammar perspective................................... 82 5.4.1 Complex expressions in Cognitive Grammar.................. 82 5.4.2 Two types of PreCs ......................................................... 86 5.5 Hebben and zijn: different construals........................................ 92 5.6 Conclusion................................................................................. 96 6 Constructional semantics: ParCs vs. PreCs............................. 97 6.1 Introduction ............................................................................... 97 6.2 Traversal.................................................................................... 99 6.2.1 A broad definition............................................................ 99 6.2.2 Complete traversal......................................................... 101 6.2.3 Partial traversal.............................................................. 105 6.2.4 Perimeter traversal......................................................... 113 6.3 Completely P........................................................................... 114 6.4 Result....................................................................................... 118 6.5 ParCs with verder and ‘length modifiers’ ............................... 123 6.6 Semantic differences between ParCs and PreCs ..................... 132 7 The grammatical constituency of ParCs................................. 135 7.1 Introduction ............................................................................. 135 7.2 Semantic similarities between ParCs and SCVCs................... 135 7.3 ParCs and motion verbs with direct objects ............................ 138 7.4 ParCs, PreCs, and the holistic/partitive effect......................... 145 7.5 ParCs and PreCs: different grammatical constituency ............ 149 7.5.1 ParCs and PreCs: different semantics............................ 149 7.5.2 Grammatical constituency in Cognitive Grammar........ 152 7.5.3 The grammatical constituency of ParCs and PreCs....... 154 7.6 Conclusion............................................................................... 163

8 Constituency tests revisited........................................................ 165 8.1 Introduction ............................................................................. 165 8.2 Attested data supporting a ‘particle analysis’?........................ 166 8.3 Attested data supporting a ‘postposition analysis’? ................ 173 8.4 The significance of the data from the constituency tests......... 179 8.4.1 A cognitive-grammar perspective ................................. 179 8.4.2 The significance of the data presented in 8.2 ................ 182 8.4.3 The significance of the data presented in 8.3 ................ 186 8.5 Conclusion............................................................................... 193 9 Conclusion...................................................................................... 195 9.1 A summary of the analysis ...................................................... 195 9.2. General conclusions ................................................................ 197 9.3 Suggestions for further research.............................................. 199 References...................................................................................... 205 Sources............................................................................................ 213 Samenvatting in het Nederlands............................................... 215 Curriculum vitae .......................................................................... 221

Chapter 1 Introduction 1.1 A long-standing problem in Dutch linguistics This study offers a new perspective on a long-standing problem in Dutch linguistics: what is the grammatical constituency of constructions such as (1)-(3)? (1) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove)1 a girl cycles the Doctor Larij-road on/up ‘A girl is cycling onto the Doctor Larij Road’ (2) de sleepboot [kruipt ...] de kant op (jnlmay94) the tow-boat creeps the bank on/up ‘The tow boat is creeping up the bank’ (3) hun fietspontje [vaart] het kanaal over (jgdapr95)

their bicycle-ferry sails the canal over ‘Their bicycle ferry sails across the canal’

The constructions, referred to as Particle Constructions (ParCs) in this study, consist of a subject nominal, a verb of motion, a second nominal, and an adposition.2 One possibility is that their grammatical constituency is similar to that of constructions such as (4)-(6), referred to as Preposition Constructions (PreCs) in this study. (4) Een jongen [… fietste] op de Rijksparallelweg (mcnov93ove) a boy cycled on the national-parallel-road ‘A boy was cycling on the National Parallel Road’ (5) [Ze] klom [...] op de kant (gp95-2) ‘She climbed onto the bank’ (6) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrock-canal ‘We are sailing along the Haren-Rüttenbrock Canal’

1 Examples followed by such combinations of letters and numbers between brackets are from the ‘38 million words corpus’ of the Institute for Dutch Lexicology (http://www.inl.nl). 2 The term ‘adposition’ is used as a general term for prepositions, postpositions, and particles.

12 Chapter 1 They too consist of a subject nominal, a verb of motion, a second nominal, and an adposition, but in PreCs the adposition precedes, rather than follows the non-subject nominal. In these constructions, the adposition is a preposition, which forms a constituent, a prepositional phrase (prePP), with the nominal that follows it. One possible analysis of ParCs is that the adposition is a postposition, which forms a constituent, a postpositional phrase (postPP), with the nominal that precedes it.

An alternative possibility is that the grammatical constituency of ParCs is more like that of constructions such as (7) and (8), which contain a separable complex verb (SCV) and are referred to as SCV Constructions (SCVCs) in this study:3 (7) Iedereen dronk zijn drankje op (mcmar95ove) everyone drank his drink up ‘Everyone finished their drinks’ (8) de gevangene haalt de trekker over (jnlfeb93) the prisoner pulls the trigger over ‘The prisoner pulls the trigger’ SCVs (see e.g. de Vries 1975, de Haas and Trommelen 1993, Haeseryn et al. 1997, Booij 1998, Blom 2005) are called ‘complex’ because they consist of a verb and an adposition, or particle; they are called ‘separable’ because the particle can occur as a separate word from the verb, as in (7) and (8), but may also directly precede the verb, in which case they are written together as one word: (9) Aleksey had het opgedronken (wk199417) Aleksey had it up-drunk ‘Aleksey had finished it’ (10) dat het slachtoffer [...] de trekker heeft overgehaald (mcfeb93ove) that the victim the trigger has over-pulled ‘that the victim has pulled the trigger’ Notice that SCVCs too consist of a subject nominal, a verb, a second nominal, and an adposition. The alternative analysis of ParCs is, in other words, that the adposition is a particle, which forms a constituent, an SCV,

3 There are also analyses of these constructions in which the non-subject nominal and the adposition are argued to form a small clause (see Blom 2005: 46-52 for a discussion of such analyses). Since that type of analysis has not, as far as I am aware, been proposed for ParCs, it will not be considered in this study.

Introduction 13 with the verb, and that the non-subject nominal is the direct object of that SCV.

Earlier analyses have not, in my view, resolved the issue. They rely on constituency tests such as passivization, topicalization, and auxiliary choice to determine the grammatical constituency of ParCs. Grammaticality judgments diverge, the tests produce conflicting results, and there is disagreement about the validity of some of the tests. Some linguists think, for example, that ParCs do not have passive counterparts and conclude that the non-subject nominal cannot be the direct object of an SCV, but must be a complement of a postposition. Others agree that ParCs do not have passive counterparts, but do not consider that a valid argument against an SCV-with-direct-object analysis; they refer to other verbs that quite clearly take a direct object, but which do not have passive counterparts. Yet others, finally, consider passive ParCs to be grammatical, from which they conclude that the adposition cannot be a postposition. This state of affairs warrants a different approach. 1.2 A different approach The title of the dissertation sums up the three essential ingredients of the analysis: constructions, constraints, and construal. They are discussed in this section.

The present study considers the question of the grammatical constituency of ParCs to be intimately related to another old problem associated with ParCs: how do they differ semantically from PreCs? Here too, I do not think that previous studies have provided a satisfactory answer. The Dutch reference grammar Algemene Nederlandse Spraakkunst (Haeseryn et al. 1997: 525) notes, for example, that it is, at the time of their writing, not yet possible to state clear rules for the use of the two constructions. Helmantel (2002), the most extensive study of the semantic differences between ParCs and PreCs, fills some of this hiatus, but still leaves a number of questions unanswered.

While previous studies treat the constituency question and the semantic differences between ParCs and PreCs as independent issues, the present study hypothesizes that the key to the grammatical constituency of ParCs lies in an analysis of the semantic differences between ParCs and PreCs. The theoretical basis for this hypothesis is Cognitive Grammar (Langacker 1987, 1991a, 1991b, 2000), in which grammatical constituency is defined as “the order in which simpler symbolic structures are successively integrated to form progressively more elaborate ones” (Langacker 2000: 149). What makes a comparison of ParCs and PreCs

14 Chapter 1 particularly interesting is that, as was noted in section 1.1, they consist of the same components: a subject nominal, a verb of motion, a second nominal, and an adposition. The present study hypothesizes that the semantic differences between the two constructions can be accounted for in terms of the different ways in which these components combine, i.e. in terms of their different grammatical constituency.

This hypothesis includes the idea that the semantic structure of an adposition is the same in both constructions. The dissertation explores the idea of interpretation as constraint satisfaction (Fauconnier 1990, 1994, 1997, Verhagen 1997), i.e. the idea that the semantic structure of a language expression imposes a (set of) constraint(s) on its possible interpretations. An optimal interpretation of the expression satisfies both this (set of) constraint(s) and the constraints imposed by the context in which the expression occurs. This, of course, is fully in line with the basic cognitive-linguistic tenet that language expressions highly underdetermine the rich interpretations that are associated with them (e.g. Fauconnier 1994, 1997, Langacker 1987); it is, as I see it, a more specific version of it. It does, however, constitute a departure from another basic assumption in Cognitive Linguistics, namely that polysemy is the norm in both lexical items and constructions.

A second basic tenet of Cognitive Linguistics is that semantic structure is conceptualization: the meaning of a word or construction is not a referent in the world, but a concept, or more precisely, ‘a particular pattern of neurological activity’ (Langacker 1987: 100). The present study supports this view: such a view is considered essential for a semantic analysis of ParCs and PreCs. Whether or not, for example, a particular motion event can be described by means of a certain ParC turns out to depend in part on whether the language user considers the mover to be headed towards a certain result. Construal (see e.g. Langacker 1987: 138), i.e. the particular way of conceptualizing a certain event, is therefore of essential importance.

The data on which the analysis is based, finally, are a large number of authentic examples. The majority of them were gathered from the ‘38 million words corpus’ of the Institute for Dutch Lexicology (INL), but they are supplemented (i) by examples that I just happened to come across, in newspapers and novels, on radio and TV, and in personal conversations, and (ii) by examples from the Internet that were obtained through search engine Google. The use of authentic data not only minimizes possible disagreement about their grammaticality; it also reduces the risk of overlooking generalizations due to considering only typical instances of a construction.

Introduction 15 1.3 Overview of the study The study is organized as follows. Chapter 2 presents an overview of the constituency tests that previous studies have used to determine the grammatical constituency. The picture that emerges is a messy one: different studies use different sets of tests, the tests provide conflicting results, and there is disagreement about not only the grammaticality of the data involved, but also the validity of some of the tests. The chapter concludes that a different approach is warranted.

The different approach includes a semantic comparison of ParCs and PreCs, which is why Chapter 3 discusses earlier semantic studies that contrast the two constructions, the most comprehensive of which is Helmantel (2002). The chapter identifies a number of problems and questions that remain unanswered.

One of these is: what is the semantics of the specific adpositions that occur in the constructions? This question is addressed in Chapter 4 in the form of a case study: it explores to what extent the uses of Dutch over as a preposition can be accounted for in terms of a single semantic constraint. It argues that earlier cognitive-linguistic analyses of over in Dutch (and English) posit too many senses, because they overlook its force-dynamic aspects.

Chapter 5 focuses on PreCs, i.e. constructions such as (4)-(6), which consist of a subject nominal, a verb of motion, and a prePP. So far I have been talking about them as though they constitute a single construction, but they are, in fact, generally classified into two types (unergative and unaccusative), depending on the role of the prePP (location vs. endpoint). The auxiliaries hebben ‘have’ and zijn ‘be’ are considered to be diagnostic for the two constructions. One of the problems identified in Chapter 3, however, is that closer inspection of actual language use reveals that hebben and zijn in fact cross-cut the proposed distinction. Another problem is that ‘location’ vs. ‘endpoint’ do not appear to be the right characterizations in the case of prePPs with path adpositions, such as door ‘through’, langs ‘along’, or over ‘over’. The chapter concludes (i) that two types of PreCs should be distinguished on the basis of the way in which the prePP integrates semantically with the verb, i.e. as a modifier or as a complement, (ii) that most PreCs with path adpositions are complements, and (iii) that hebben and zijn are not diagnostic for these two modes of integration, but cross-cut the distinction, imposing their own construal on the motion event (type of activity vs. change of location).

16 Chapter 1

Chapter 6 analyzes ParCs in terms of a single semantic constraint: a ParC designates a motion event in which a trajector traverses4 a landmark so that the result is that the trajector is completely P (i.e., completely in, on, through, etc. the landmark). The chapter shows that the wide variety of interpretations that ParCs allow can be explained in terms of (i) this constructional constraint, (ii) the semantic constraints imposed by the words that make up the construction, and (iii) constraints imposed by the context in which the construction occurs. The chapter furthermore shows that it is the three notions in italics, traversal, result, and completely P, that make ParCs essentially different from PreCs: ParCs and PreCs both designate motion events involving a trajector and a landmark, but PreCs do not require traversal, are not necessarily resultative, and do not require that the trajector ends up completely P.

Chapter 7 argues that the semantic differences between ParCs and PreCs can be accounted for in terms of differences in their grammatical constituency. This chapter provides, in other words, an answer to the long-standing question of the grammatical constituency of ParCs. It proposes that while the adposition of a PreC forms a constituent (a prePP) with the nominal that follows it, the adposition and verb form a SCV which takes a subject and a certain type of direct object. This proposal is based on three types of semantic similarities: (i) similarities between ParCs and SCVCs, (ii) similarities between ParCs and constructions from other languages in which a motion verb takes a direct object, and (iii) similarities between the semantic contrast between ParCs and PreCs and the ‘holistic/partitive contrast’ (e.g. Levin 1993) in ‘argument/oblique alternations’ (Beavers 2006). Chapter 8 revisits the constituency tests that figured in earlier analyses of the grammatical constituency of ParCs. It first of all shows that inspection of actual language use reveals that quite a number of constructions that (some) earlier studies deemed impossible are in fact possible, which underscores the importance of using authentic material. It then goes on to argue that a positive grammaticality judgment about one construction does not per se warrant a conclusion about the grammatical constituency of another construction. Such a conclusion is only possible after careful consideration of the semantics involved: if the constructions are semantically similar, they can be argued to have the same grammatical constituency. Chapter 9 concludes and suggests avenues for further research.

4 Traversal is defined in a special way in this study, namely as ‘changing location where the landmark is’, see section 6.2.

Chapter 2 Postposition or particle: Constituency tests 2.1 Introduction Earlier studies that have addressed the question of the grammatical constituency of ParCs arrive at different conclusions. Some (Paardekooper 1959, Helmantel 2002) conclude that ParCs are syntactically similar to PreCs and analyze the adposition of a ParC as a postposition, which forms a postPP with the nominal that precedes it. Other studies (Verkuyl and Zwarts 1992, Beeken 1993) conclude that ParCs are syntactically similar to SCVCs and analyze the adposition of a ParC as a particle, which forms an SCV with the verb and this SCV takes a subject and a direct object. Most studies (de Schutter 1974, van Riemsdijk 1978, Hoekstra 1984, Luif 1992, de Haas and Trommelen 1993, Blom 2005), however, conclude that ParCs show ambivalent behavior: they are like PreCs in some respects and like SCVCs in others. The ambivalent behavior is accounted for in terms of structural ambiguity: the adposition is essentially a postposition, but language users may reanalyze the structure so that the adposition becomes a particle that forms an SCV with the verb and the non-subject nominal a direct object.

These different conclusions are not due to different methodologies: all of the studies rely on constituency tests such as, for example, passivization, topicalization, and auxiliary choice to determine the grammatical constituency of ParCs.1 By presenting an overview of the tests and their different outcomes in different studies, this chapter identifies four reasons why the studies have come to different conclusions: (i) they do not use the same set of tests, (ii) some tests are believed to yield conflicting results, (iii) there is disagreement about the grammaticality of the examples, and (iv) there is disagreement about the validity of some of the tests. Sections 2.2-2.10 each discuss a different test; section 2.11 concludes that the results warrant a new study that is based on a different methodology.

1 Verkuyl and Zwarts (1992) is an exception. They posit a complex predicate analysis of ParCs on the basis of similar ‘aspectual behavior’ of ParCs and SCVCs: they can have durative and terminative interpretations. Verkuyl and Zwarts then support their proposal by a number of the constituency tests that are discussed in the present chapter.

18 Chapter 2

2.2 No verb to form an SCV with A compelling argument for a ‘postposition analysis’ is that nominal-adposition (NP-P) sequences occur in constructions that do not contain a verb at all, or that do not contain a verb with which the adposition could form an SCV (de Schutter 1974, van Riemsdijk 1978, Luif 1992, Helmantel 2002, Cloutier 2006); instead, the NP-P sequence occurs in a constructional slot in which prePPs can also occur, see e.g. (1)-(3) from van Riemsdijk (1978: 94-97) and (4) from de Schutter (1974: 284): (1) De gevangenis in met die boef!

the jail in with that rascal ‘To jail with that rascal!’

(2) de weg de stad in the road the city in ‘the road into the city’ (3) omdat hij absoluut wilde meerijden de berg af because he absolutely wanted with-drive the mountain off/down ‘because he absolutely wanted to ride along down the mountain’ (4) Vrij ver het bos in vind je dan plots zo’n huis quite far the wood in find you then suddenly such-a house ‘Quite far into the wood you suddenly then find such a house’ Helmantel (2002), for example, draws her conclusion that NP-P sequences are postPPs on the basis of data such as these. (1) and (2) do not contain a verb at all. (3) and (4) do contain verbs, but in neither case can the adposition of the NP-P sequence be said to form an SCV with it. The verb meerijden ‘ride along, be given a ride’ in (3) is itself an SCV: af cannot form an SCV with rijden ‘ride, drive’, because rijden forms an SCV with mee ‘with, along’. Nor does in in (4) form an SCV with the verb vind ‘find’; vind takes as its direct object zo’n huis ‘such a house’ and describes an event of which the phrase vrij ver het bos in ‘quite far into the wood’ specifies the location. 2.3 Non-subject nominal and adposition not adjacent While the NP-P sequences in (1)-(4) show a similar distribution to that of prePPs, this section focuses on a distributional difference between them. Consider, for instance, (5) from Haeseryn et al. (1997: 508).

Postposition or particle: Constituency tests 19

(5) Hij rijdt de garage in he drives the garage in ‘He is driving into the garage’ There is general agreement that a nominal such as de garage ‘the garage’ and an adposition such as in need not be adjacent to one another: modifiers or auxiliary verbs may occur between these elements (de Schutter 1974, van Riemsdijk 1978, Verkuyl and Zwarts 1992, Beeken 1993, de Haas and Trommelen 1993, Haeseryn et al. 1997), see e.g. (6) from Haeseryn et al. (1997: 509) and (7) from van Riemsdijk (1978: 98). (6) Hij rijdt de garage altijd heel voorzichtig in he drives the garage always very carefully in ‘He always very carefully drives into the garage’ (7) omdat hij de boom is in geklommen because he the tree is in climbed ‘because he has climbed into the tree’

In this respect, ParCs are similar to SCVCs and different from PreCs. This is one of the reasons why Verkuyl and Zwarts (1992) and Beeken (1993) suggest that constructions such as (5) are not postpositional, but consist of a SCV with a direct object. They do not, however, take into account data such as (1)-(4). Those who do (de Schutter 1974, van Riemsdijk 1978, de Haas and Trommelen 1993, Haeseryn et al. 1997) believe that data such as (6) and (7) support the idea that while the adposition is essentially a postposition, it can be reanalyzed as the particle of an SCV.

Luif (1992) suggests that reanalysis is more likely to happen with the adpositions binnen ‘inside’ and voorbij ‘past’, than with in ‘in’ or uit ‘out’. One of his reasons for this is that in his view the NP-P sequence can be ‘separated’ by modifiers or auxiliaries in the case of binnen and voorbij, but not so easily in the case of in or uit, compare e.g. (8) and (9) from Luif (1992: 163). (8) omdat ik die kamer moet proberen binnen te sluipen because I that room have-to try inside to sneak

‘because I have to try to sneak into that room’ (9) *omdat ze de sloot moesten proberen in te springen because they the ditch had-to try in to jump ‘because they had to try to jump into the ditch’

20 Chapter 2

Luif does not, in other words, share the positive grammaticality judgments in (6) and (7). 2.4 Topicalization The tests discussed in this section and the next, topicalization and pronominalization, have been used to show that ParCs are structurally ambiguous. Topicalization is based on the assumption that a string of words that can occur in topicalized, i.e. sentence-initial, position forms a constituent. Haeseryn et al. (1997: 509), for example, provide the following examples: (10) De garage in rijdt hij altijd zelf the garage in drives he always himself ‘Into the garage he always drives himself’ (11) De garage rijdt hij altijd heel voorzichtig in the garage drives he always very carefully in ‘The garage he always drives into very carefully’ (10) shows that an NP-P sequence can occur in topicalized position, which is taken as evidence that it forms a constituent, a postPP. NP-P sequences share this distribution with prePPs, but not with the direct object and particle of an SCV. (11), on the other hand, shows that the non-subject nominal may also occur in topicalized position without the adposition, which is taken as evidence that the nominal and the adposition do not form a constituent, but that the adposition forms an SCV with the verb.

Haeseryn et al.’s positive grammaticality judgment in (10) (see also Paardekooper 1959, 1966, and de Vries 1975) is not shared by everyone, however. Verkuyl and Zwarts (1992) and Beeken (1993), for example, do not accept examples such as (10), see e.g. (12) from Verkuyl and Zwarts (1992: 395). This supports their idea that het bos ‘the wood’ and in do not form a postPP, but that in is a particle. (12) *?Het bos in is zij ondanks mijn advies toch gelopen the wood in is she in-spite-of my advice nevertheless walked *?‘Into the wood she has walked in spite of my advice’

Luif (1992) does not rule out the possibility of topicalization of NP-P, but thinks that the results sound rather ‘forced’. His judgments in (13) and (14) (from Luif 1992: 162) are in line with the distinction that he makes


between constructions with binnen ‘inside’ and voorbij ‘past’ and those with in ‘in’ and uit ‘out of’. (13) ?Die kamer binnen sluip ik nooit

that room inside sneak I never ?‘Into that room, I never sneak’

(14) Zijn hol uit kwam hij niet his hole out came he not ‘Out of his hole, he did not come’

Luif also mentions another type of topicalization, in which the

adposition and the verb (P-V) occur in sentence-initial position. This is possible, according to Luif, with voorbij and binnen, which suggests that they are SCVs, but not with in or uit, supporting Luif’s idea that those are postpositions, see his examples in (15) and (16) (Luif 1992: 163). (15) Voorbijlopen mag jij je buurman niet past-walk can you your neighbor not ‘You are not allowed to walk past your neighbor’ (16) *Inspringen mochten ze de sloot niet in-jump could they the ditch not ‘They were not allowed to jump into the ditch’ Blom (2005: 119), however, rules out the possibility of topicalization of P-V altogether, which supports her conclusion that adpositions in what I call ParCs are postpositions. 2.5 Pronominalization Another test that has been taken to show that ParCs are structurally ambiguous is pronominalization (van Riemsdijk 1978, Hoekstra 1984, de Haas and Trommelen 1993). In the case of nominals there are two kinds of proforms in Dutch. So-called ‘r-pronouns’ such as waar ‘where, that’ replace the NP-complement of a preposition. ‘Non-r-pronouns’ such as die ‘who, that [non-neuter]’ or dat ‘that [neuter]’ replace ‘bare’ nominals, i.e. subjects and objects.

(17), for example, contains an r-pronoun, the relative pronoun waar ‘where, that’. It occurs at the beginning of a relative clause that corresponds to the main clause in (18): waar refers to the nominal de tafel ‘the table’, which is the complement of the preposition aan ‘at’. A non-r-pronoun, such as die ‘who, that’, is not possible.

22 Chapter 2

(17) Dat is de tafel waar/*die hij vaak aan zit. that is the table where/*that he often at sits ‘That is the table that he often sits at’

(18) Hij zit vaak aan de tafel he sits often at the table ‘He often sits at the table’

(19), on the other hand, contains a non-r-pronoun: the relative

pronoun die ‘who, that’. It occurs at the beginning of a relative clause that corresponds to the main clause in (20): die refers to the nominal zijn hond ‘his dog’, which is the direct object, a bare nominal, of the SCV uitlaten ‘walk’ (lit. ‘let out’). An r-pronoun, such as waar ‘where, that’, is not possible. (19) Dat is de hond die/*waar hij uitliet that is the dog that/*where he out-let ‘That is the dog that he walked’ (20) Hij liet zijn hond uit he let his dog out ‘He walked his dog’

ParCs have been argued to be structurally ambiguous because the non-subject nominal of a ParC, such as die brug ‘that bridge’ in (21), is considered to be replaceable by either type of pronoun, see (22) from van Riemsdijk (1978: 99). (21) Ik rij altijd die brug over I drive always that bridge over

‘I always drive across that bridge’ (22) Dat is de brug waar/die ik altijd over rijd that is the bridge where/that I always over drive

‘That is the bridge that I always drive across’ That is, it can be replaced by waar, which suggests that die brug is the complement of a postposition, but it can also be replaced by die, which suggests that die brug is a bare NP, a direct object.

De Schutter (1974), Luif (2000), and Helmantel (2002), however, object to the idea that a relative clause with waar such as the one in (22) corresponds to a main clause such as (21). They argue that when waar is used, the clause corresponds to a construction with a preposition, such as the PreC in (23), not to the ParC in (21).


(23) Ik rijd altijd over die brug I drive always over that bridge ‘I always drive across that bridge’

De Schutter concludes that data with pronouns suggest that the non-subject nominal is a direct object. Luif considers non-r-pronouns to be more likely to occur with binnen and voorbij than with in and uit, which supports his idea that the former are more susceptible to reanalysis than the latter, see (24) and (25) from Luif (1992: 164). (24) de buurman die jij voorbijliep the neighbor that you past-walked the neighbor that you passed (25) *het bed dat ik uitkroop the bed that I out-crept ‘the bed that I crept out of’ Helmantel, finally, who analyzes adpositions of ParCs as postpositions, does not consider data with non-r-pronouns. 2.6 Passivization ParCs are often assumed not to allow passivization (see Paardekooper 1966, de Schutter 1974, Verkuyl and Zwarts 1992, de Haas and Trommelen 1993, Blom 2005); consider for example ParC (26) with the corresponding passive in (27), judged ungrammatical by Blom (2005: 119). (26) Hij reed de garage in he drove the garage in ‘He drove into the garage’ (27) *De garage werd in gereden the garage was in driven ‘The garage was driven into’ The conclusion that Paardekooper (1959, 1966), de Haas and Trommelen (1993), and Blom (2005) draw from these data is that constructions such as (26) are quite different from SCVCs: de garage ‘the garage’ in (26) is not a direct object, because it cannot serve as the subject of a corresponding passive.

While de Schutter (1974) and Verkuyl and Zwarts (1992) have the same intuitions with respect to the grammaticality of examples such as (27),

24 Chapter 2

they do not think that this necessarily means that ParCs and SCVCs are syntactically different. They point to several other verbs that are clearly ‘two-place predicates’, but that do not passivize. De Schutter mentions bereiken ‘reach’, verlaten ‘leave’ and naderen ‘approach’ (1974: 283), and Verkuyl and Zwarts nadoen ‘imitate’, bevallen ‘please’, and bevelen ‘order’ (1992: 398-399). They object, in other words, to the validity of using passivization as a test for direct-objecthood.

De Vries (1975) and Hoekstra (1984) take yet another position. According to them, ParCs are passivizable, see for instance the passive ParC in (28), from Hoekstra (1984: 172). (28) Dat kanaal is nog nooit door iemand over gezwommen

that canal is so-far never by anyone over swum ‘That channel has so far never been swum across by anyone’

On the basis of examples such as these, de Vries and Hoekstra argue that the adposition of such constructions is not a postposition. 2.7 Auxiliary choice Similarly diverging positions can be found with respect to data with the perfect auxiliaries hebben ‘have’ and zijn ‘be’. De Haas and Trommelen (1993), for example, argue that the non-subject nominal of a ParC cannot be a direct object, because ParCs take zijn, see (29) from de Haas and Trommelen (1993: 105), while transitive verbs in Dutch usually take hebben. (29) omdat hij het land is ingelopen

because he the land is in-walked ‘because he has walked onto the land’

Paardekooper (1959: 315) argues for the same reason that such constructions must consist of a postposition.

De Schutter (1974), however, dismisses the use of hebben as a test for objecthood. According to him, there are several ‘two-place predicates’ that take the perfect auxiliary zijn. Hoekstra (1984), Luif (1992), and Blom (2005), on the other hand, accept hebben as a test for objecthood, but observe that hebben ‘to have’ does sometimes occur in ParCs, cf. e.g. (30) from Luif (1992: 165).


(30) Ik heb/ben de rivier overgezwommen I have/am the river over-swum ‘I have swum across the river’

They analyze the adposition as a postposition in the case of zijn, but as a particle in the case of hebben: the possibility of using hebben is taken as evidence that reanalysis of the constituent structure has taken place. 2.8 Nominalization Nominalization is another test where we find diverging grammaticality judgments. Blom (2005) considers nominalization to be impossible in the case of ParCs, cf. the asterisk in (31) from Blom (2005: 119), which supports her idea that the adposition in such constructions is a postposition. (31) *het in rijden van de garage the in drive of the garage ‘the driving into of the garage’ Luif (1992) thinks that while ParCs with in and uit ‘out (of)’ cannot be nominalized, ParCs with voorbij ‘past’ and binnen ‘inside’ can, cf. (32) from Luif (1992: 163), which supports his idea that constructions with voorbij and binnen can be reanalyzed as constructions with an SCV and a direct object. (32) het binnensluipen van die kamer the inside-sneak of that room ‘sneaking into that room’ Verkuyl and Zwarts (1992) and Beeken (1993), finally, do not make such qualifications; they think that ParCs can be nominalized, which supports their idea that they consist of an SCV that takes a direct object. 2.9 Caused motion An argument for a postposition analysis of ParCs is based on the existence of constructions such as (33), from Luif (1992: 160), alongside ParCs such as (34).

26 Chapter 2

(33) Ik rij de kar het huis in I drive the cart the house in ‘I am driving the cart into the house’

(34) Ik rij het huis in I drive the house in ‘I am driving into the house’

Notice that (33) contains three nominals and expresses caused motion: because of what the subject referent does, the referent of the second nominal moves with respect to the referent of the third nominal. Luif (1992: 160), de Haas and Trommelen (1993: 105), and Blom (2005: 272) argue as follows: since the second nominal (de kar ‘the cart’) is the direct object of (33), the third nominal (het huis ‘the house’) cannot be; the NP-P sequence het huis in ‘into the house’ forms therefore a postPP.

Hoekstra (1984), however, does not seem to rule out the possibility that the nominal preceding the adposition is an object in such constructions. He mentions a similar example, see (35) from Hoekstra (1984: 172), about which he states that it consists of “a direct object (het schip), and a NP P sequence, the structure of which may be ambiguous”. (35) dat de kapitein het schip de haven binnen heeft gevaren that the captain the ship the harbor inside has sailed ‘that the captain has navigated the ship into the harbor’ It seems to be the case, in other words, that Hoekstra leaves open the possibility that de haven ‘the harbour’ is a direct object too. 2.10 Coordination A final constituency test involves coordination. Verkuyl and Zwarts (1992: 394) and Beeken (1993: 427) argue that ParCs and SCVCs show the same behavior with respect to coordination. Neither the coordinated SCVCs in (36) nor the coordinated ParCs in (37) allows the occurrence of the adposition at the end of the sentence (examples from Verkuyl and Zwarts 1992: 394). (36) omdat Henk de auto inrijdt en Frans de motor (* in) because Henk the car in-drives and Frans de motorbike in ‘because Henk runs in the car and Frans the motorbike’


(37) omdat Henk het bos inloopt en Frans het park (* in) because Henk the wood in-walks and Frans the park in ‘because Henk enters the wood and Frans the park’ Corver (1990), on the other hand, gives examples in which NP-P and NP-P are coordinated, see (38) from Corver (1990: 301). (38) Dat Jan niet alleen de berg op

that Jan not only the mountain up maar ook de berg af moest rijden

but also the mountain down had-to ride ‘that Jan did not just have to ride up the mountain, but also down the mountain’

Corver considers this support for a postposition analysis, because he assumes that only constituents can be coordinated. Beeken (1993: 429, fn. 7) objects that this type of coordination is found in SCVCs too, see her example in (39). (39) Hij zou achtereenvolgens moeder op en vader af-bellen.

he would successively mother up and father off-call ‘He would successively call mother up and father off’

2.11 Conclusion The chapter has described four reasons for the current disagreement about the grammatical constituency of ParCs: (i) the existing studies do not all use the same set of constituency tests, (ii) some tests are believed to yield conflicting results, (iii) there is disagreement about the grammaticality of the examples, and (iv) there is disagreement about the validity of some of the tests.

I conclude that since earlier studies leave the constituency question unresolved, a new approach to the problem is warranted. One way in which the present study is different from the ones discussed in this chapter is that it takes a cognitive-grammar perspective: it assumes that “symbolic considerations are critical to the determination of grammatical constituency” (Langacker 1987: 366). The backbone of the present study is therefore a semantic analysis of ParCs, which is arrived at by contrasting ParCs and PreCs. Chapter 3 discusses earlier studies of the semantic differences between ParCs and PreCs.

28 Chapter 2

A second way in which the present study is different is that it is based on attested, rather than constructed examples. After I have presented my analysis of the semantics and grammatical constituency of ParCs in Chapters 6 and 7, Chapter 8 revisits the constituency tests discussed in this chapter, checking the grammaticality judgments against actual language usage and evaluating the validity of the tests for determining grammatical constituency.

Chapter 3 Previous semantic studies 3.1 Introduction Since this study hypothesizes that the grammatical constituency of ParCs can be determined on the basis of a semantic comparison of ParCs and PreCs, the present chapter discusses previous studies that have addressed the semantic differences between the two constructions. Section 3.2 presents Kraak and Klooster’s (1968) characterization of the difference in terms of imperfective vs. perfective aspect and the objections that it has evoked. Section 3.3 gives an overview of Helmantel’s (2002) analysis in terms of direction vs. location; it is the most extensive study of the semantic differences to date. Section 3.4 identifies a number of problems for Helmantel’s analysis. Section 3.5 briefly states how these problems will be taken up in this study. 3.2 Perfective and imperfective A good starting point for an analysis of the semantic differences between ParCs and PreCs is Kraak and Klooster’s (1968) characterization of the difference between the ParC in (1) and the PreC in (2), from Kraak and Klooster (1968: 223):1 (1) De man klimt de ladder op the man climbs the ladder on/up ‘The man is climbing onto/up the ladder’ (2) De man klimt op de ladder the man climbs on the ladder ‘The man climbs/is climbing onto the ladder’ According to Kraak and Klooster, (1) requires that the man moves upward on the ladder, climbing it using its rungs: the ladder has to be in an upright position. In (2), on the other hand, the ladder can have any orientation (it could be lying flat on the floor); the sentence conveys only that the man is at some point op de ladder ‘on the ladder’ as the result of his climbing. The difference between (1) and (2) is defined by Kraak and Klooster in terms of a difference in aspect: PreC (2) is perfective, because it involves

1 Glosses and translations in (1)-(6) are mine, MB.

30 Chapter 3

a moment of completion (when the man is on the ladder); ParC (1) is imperfective, because there is no such moment of completion. They use the temporal modifier een uur (lang) ‘for an hour’, which is compatible with imperfectives but not with perfectives, to support their idea that ParCs with op ‘on/up’, in ‘in’, and uit ‘out (of)’ are imperfective, see their example (3) with in (1968: 223): (3) We zijn een uur (lang) het bos ingelopen we are an hour (long) the wood in-walked ‘We have been walking into the wood for an hour’ Kraak and Klooster’s analysis is restricted to PreCs and ParCs with op, in, and uit: they do not think that there is such an aspectual difference between PreCs and ParCs with door ‘through’, langs ‘along’, and over ‘over, across’ (1968: 224). Schermer-Vermeer (1970) agrees with Kraak and Klooster’s analysis that PreC (2) is perfective and that as a result of (2), the man is on the ladder, but does not think that ParCs are imperfective. In her view, (3) is not grammatical (see also Helmantel 2002: 70), which supports Schermer-Vermeer’s idea that ParC (3) is perfective. She uses the temporal modifier in een minuut ‘in a minute’, which is compatible with perfectives but not with imperfectives, to support her idea that ParCs are perfective (see also Helmantel 2002: 71, Verkuyl and Zwarts 1992: 385): (4) De man is in een minuut de ladder opgeklommen the man is in a minute the ladder on/up-climbed ‘The man climbed up the ladder in a minute’ Verkuyl and Zwarts (1992) add some further subtleties. They consider examples such as (3) grammatical, but only under an iterative interpretation (the man walks into the wood again and again), which is further support of the perfectivity (or ‘terminativity’, to use their term) of ParCs. Another observation is that ParCs may contain verder ‘further’, see (5), which ‘typically terminative’ constructions do not, see (6), both from Verkuyl and Zwarts (1992: 385-386). (5) Piet is het bos verder ingelopen Piet is the wood further in-walked ‘Piet walked further into the wood’

Previous semantic studies 31

(6) ?Piet is verder vertrokken Piet is further left ?‘Piet left further’ They conclude that ParCs can receive terminative and durative, i.e. imperfective, interpretations (1992: 386). Helmantel (2002), finally, dismisses the possibility of analyzing the difference between PreCs and ParCs in terms of aspect altogether. She notes that PreCs such as (2) can be imperfective too, namely when the PP conveys not the endpoint, but the location of the motion (see also Schermer-Vermeer 1970: 97), cf. (7), from Helmantel (2002: 71): (7) We hebben een uur in het bos gelopen we have an hour in the forest walked ‘We have been walking in the forest for an hour’ She therefore concludes that the difference between PreCs and ParCs cannot be captured in terms of aspect. 3.3 Location and direction What Helmantel (2002) proposes instead is an analysis in terms of location vs. direction, which are notions that have, in fact, been used before to describe the difference between PreCs and ParCs.2 Haeseryn et al. (1997), for example, state that (what they call) postpositions convey the direction of the motion described by the verb, while prepositions may convey the direction or the location of the motion (1997: 524, see also den Hertog 1973, Schermer-Vermeer 1970).

What is new in Helmantel’s analysis, however, is that she considers only postpositional phrases (postPPs)3 to be directional: prepositional phrases (prePPs) are not.4 She defines ‘directionality’ as ‘change of location’

2 Helmantel herself refers to van Riemsdijk (1978) and Koopman (1997). 3 In this discussion of Helmantel’s semantic analysis, I use her terms ‘postposition’ and ‘postPP’ to refer to the adposition of a ParC and the combination of such an adposition and the nominal that precedes it. 4 PrePPs with naar ‘to’, van ‘from’, and via ‘via’ form an exception; these adpositions are, according to Helmantel, inherently directional. Because they occur only in PreCs, not in ParCs, they fall outside the scope of the present study.

32 Chapter 3

(2002: 8), which is indeed involved in example (8) from Helmantel (2002: 71).5 (8) Piet fietst de berg op Piet cycles the mountain on/up ‘Piet is cycling up the mountain’ PrePPs, on the other hand, are considered to be purely locative, which is quite obviously so in a case such as (9), from Helmantel (2002: 71), but not in cases such as (10) and (11), from Helmantel (2002: 12, 17): the latter do involve change of location. She argues, however, that the directionality is not due to the prePPs, for reasons that I will address in a moment. (9) Jan woont op de berg ‘Jan lives on the mountain’ (10) Jan rent langs de snelweg Jan runs along the highway ‘Jan is running along the highway’ (11) dat Jan in de sloot gesprongen is that Jan in the ditch jumped is ‘that Jan jumped into the ditch’

Before I do, however, I want to point out a second innovative aspect of Helmantel’s analysis, which is that she classifies the adpositions that can occur in both PreCs and ParCs into two semantic types, see Table 1. Extended locative adpositions, which I shall refer to as path adpositions, have “a path specification in their function-argument structure” (2002: 11), while narrow locative adpositions, which I shall refer to as point adpositions, do not.6 This classification allows her to draw a number of semantic generalizations about PreCs and ParCs.

One such generalization is that a prePP with a point adposition denotes a point-like location, while a prePP with a path adposition denotes an extended location. Helmantel uses English prePPs to illustrate the semantic difference. The prePP in the store, with point adposition in, “denotes a point in space; no claims are made concerning a particular location of this point on a path” (2002: 11). The prePP along the river, with path adposition along, denotes a location with a certain extension; the 5 The glosses and/or translations of the quoted examples in this section have occasionally been slightly adapted. 6 I have argued elsewhere, see Beliën (2006), that uit ‘out (of)’ is not a path adposition, but a point adposition.


nominal that follows along, i.e. the river, “is linked to more than one point on the path” (2002: 11). Table 1. Helmantel’s classification Narrow locative (point) adpositions Extended locative (path) adpositions binnen ‘inside’ door ‘through’ in ‘in’ langs ‘along’ op ‘on’ om ‘around’ over ‘over, across’ rond ‘around’ uit ‘out (of)’ voorbij ‘past’

We can now return to Helmantel’s claims that the prePPs in (10) and

(11) are purely locative, and that the directionality of the examples is due to other factors. She argues that the prePP in (10), with path adposition langs, denotes an extended location, by pointing to example (12), from Helmantel (2002: 12). (12) De palen staan langs de snelweg ‘The posts stand along the highway’ Her argument runs as follows: since there is no directionality, i.e. ‘change of location’, in (12), in which langs de snelweg ‘along the highway’ combines with the stative verb staan ‘stand’, the directionality of (10) must be due to the verb rent ‘runs’. The prePP is purely locative in both (10) and (12), and it is the verb rent that ‘brings about’ the directional interpretation (Helmantel 2002: 12). The directional interpretation of (11) is ascribed not to the prePP or the verb, but to ‘the entire configuration’ (Helmantel 2002: 17). Helmantel argues that the prePP in de sloot ‘in the ditch’ fulfils a different function in (11) from that in (13) (cf. Hoekstra 1984): it is predicative in (11) and adverbial in (13), from Helmantel (2002: 18). (13) dat Jan in de sloot gesprongen heeft that Jan in the ditch jumped has ‘that Jan has been jumping in the ditch’ Helmantel gives two tests for distinguishing the two types of PPs. The first is (perfect) auxiliary choice: adverbial PPs occur with hebben ‘to have’, see (13), while predicative PPs in such constructions as (11) occur with zijn ‘to

34 Chapter 3

be’. The second test is ‘extraposition’: adverbial PPs can ‘extrapose’, i.e. occur after the verb cluster in a subordinate clause, see (14), while predicative PPs cannot, see (15), both from Helmantel (2002: 17-18). (14) dat Jan gespeeld heeft in de sloot that Jan played has in the ditch ‘that Jan has played in the ditch’ (15) *dat Jan geklommen is op de tafel

that Jan climbed is on the table ‘that Jan has climbed onto the table’

Semantically, the prePP in de sloot ‘in the ditch’ denotes a point-like

location regardless of its function. In (13) the point-like location is the location where the motion takes place. In (11) it is the endpoint of the motion: “there is a jumping (movement) event followed by the resultant state [Jan in the ditch]” (Helmantel 2002: 17).7

Let us now turn to postPPs, which are considered to be directional in Helmantel’s analysis: they “denote a process along a path” (2002: 110). She claims that the adposition in a postPP receives a directional interpretation and the nominal that precedes the adposition receives a 1-dimensional interpretation: the nominal “is mapped onto the spatial path provided by the adposition” (2002: 73). De ladder ‘the ladder’ in (16), from Helmantel (2002: 73), for example, “functions as a path along which the climbing takes place” (2002: 73). (16) De man is de ladder op geklommen

the man is the ladder on/up climbed ‘The man has climbed up the ladder’

Since I consider the latter generalization, i.e. that the nominal that precedes the adposition “denotes a path” (Helmantel 2002: 110), to be problematic (see also Gehrke 2007), I will come back to it in section 3.4. Helmantel herself, in fact, identifies a ‘small set’ of postPPs in which the nominal is not interpreted as a path, such as, for example, (17), from Helmantel (2002: 74).

7 See for other purely locative analyses of Dutch point adpositions, such as in and op, Luif (2000), Beliën (2002), and Gehrke (2007).


(17) Peter rent de klapdeuren door Peter runs the swinging-doors through ‘Peter runs through the swinging doors’ Instead, according to Helmantel, de klapdeuren ‘the swinging doors’ are interpreted as a ‘point on a path’ (2002: 74), which leads her to the conclusion that postPPs such as these constitute a different type of constructions from those that we have considered so far.

Her two-way classification of the adpositions allows Helmantel to draw two generalizations about postPPs. The first can be illustrated by means of (18) with the point adposition binnen ‘inside’, from Helmantel (2002: 112), and (19) with the path adposition over ‘over, across’, from Helmantel (1998: 381), and their representations in Figs. 1 and 2, from Helmantel (2002: 110). (18) De bal rolt [...] het strafschopgebied binnen the ball rolls [...] the penalty area inside ‘The ball rolls [...] into the penalty area’ (19) Jan fietst de brug over Jan cycles the bridge across ‘Jan cycles across the bridge’

These examples show that the path involved in a postPP may “‘occupy’ part of the object x [..., i.e. the referent of the nominal that precedes the adposition,] or the entire object” (Helmantel 2002: 110). The generalization that Helmantel offers is that in postPPs with path adpositions, see Fig. 2, the length of the path relative to object x is fixed, that is, it “equals the length of the object” (2002: 111), while in postPPs with point adpositions, see Fig. 1, the length of the path is not fixed. 8

8 Helmantel (1998: 376) identifies uit ‘out (of)’ as an exception. A postPP with uit, which Helmantel classifies as a path adposition, involves a path that can ‘occupy’ part of object x. This is one of my reasons for proposing that uit is a point adposition, see fn. 6 of this chapter and Beliën (2006).

x

binnen (x)

Fig. 1

x

over (x)

Fig. 2

36 Chapter 3

Helmantel’s second generalization is that in the case of postPPs with point adpositions there is the implication that “at the end of the movement denoted by the verb, the subject is at location P(x), where P is given by the adposition”, while there is no such implication in the case of postPPs with path adpositions (Helmantel 2002: 111), see (20) and (21) from Helmantel (2002: 111):

(21) Piet fietst de tunnel in → Piet bevindt zich in de tunnel Piet cycles the tunnel in → Piet finds REFL in the tunnel

‘Piet cycles into the tunnel (and after that: Piet is in the tunnel)’ (22) Piet fietst de tunnel door *→ Piet bevindt zich door de tunnel

Piet cycles the tunnel through *→ Piet finds refl through the tunnel ‘Piet cycles through the tunnel *(and after that: Piet finds himself through the tunnel)’

These examples reflect Helmantel’s idea that the implication is there with a postPP with in, a point adposition, see (21), but not with door, a path adposition, see the asterisk in (22). In summary then, Helmantel identifies the following differences between prePPs and postPPs. PrePPs are purely locative: they denote a point-like location (in the case of point adpositions) or an extended location (in the case of path adpositions), which constitutes either the location where the motion takes place (in the case of adverbial PPs) or the endpoint of the motion (in the case of predicative PPs). PostPPs are directional: they denote a process along a path. In the case of point adpositions, the path relative to ‘object x’ is not fixed and there is the implication that at the end of the motion, the subject referent is P(x). In the case of path adpositions, the path relative to ‘object x’ is fixed and there is no such implication. 3.4 Four problems This section discusses four problems for Helmantel’s analysis. The first concerns the implication that the subject referent is at P(x) after the motion in the case of postPPs with point adpositions, but not in the case of postPPs with path adpositions. The second concerns predicative prePPs with path adpositions. The third is a problem not only for Helmantel’s analysis, but also for other analyses that somehow relate auxiliary choice to two types of PPs: attested data show that hebben and zijn cross-cut the proposed distinctions. The fourth concerns the maximal length of the path described by a postPP with a point adposition.


First of all, I do not think that Helmantel’s claim is correct that in cases of postPPs with path adpositions, there is no implication that the subject is P(x) at the end of the motion. If we take zich bevindt ‘is located (lit. finds himself)’, which admittedly does sound odd,9 and replace it by is ‘is’, the result is perfectly acceptable: (23) Piet fietst de tunnel door → Piet is door de tunnel

Piet cycles the tunnel through → Piet is through the tunnel ‘Piet cycles through the tunnel (and after that: Piet is through the tunnel)’

This I consider a very promising result, because it suggests that a more encompassing generalization can be formulated: in the case of postPPs, with whatever type of adposition, point or path, there is the implication that at the end of the movement, the subject is P(x). An additional bonus is that this generalization also seems to apply to the subset of postPPs that Helmantel is forced to exclude from her general analysis: those in which the nominal cannot be interpreted as a path, such as (17), repeated here, for which we can formulate the implication in (24), or (25) and the implication in (26).10 (17) Peter rent de klapdeuren door Peter runs the swinging-doors through ‘Peter runs through the swinging doors’ (24) → Peter is door de klapdeuren (and after that:) ‘Peter is through the swinging doors’ (25) De jongens vluchtten [...] de grens over (jgdmay92) the boys fled the border over ‘The boys fled across the border’ (26) → De jongens zijn over de grens (and after that:) ‘The boys are across the border’ If this is going in the right direction, we have to dismiss Helmantel’s generalization that the nominal of the postPP is interpreted as a path. 9 It sounds as though Piet is some sort of elongated object that has pierced through the tunnel, or as though he is somehow scattered throughout the tunnel, just as the asbestos in (i) (http://www.aspbv.nl/toepassingen.htm, March 2008). (i) Het asbest bevindt zich door het hele materiaal the asbestos finds REFL through the entire material ‘The asbestos is located throughout the material’ 10 In the case of (26), the implication can be formulated by means of zich bevinden:: (i) De jongens bevinden zich over de grens ‘The boys are located across the border’

38 Chapter 3

Helmantel herself identified constructions such as (17) as problems for this claim, but there are more examples for which it does not sound like the right generalization. (18) De bal rolt [...] het strafschopgebied binnen the ball rolls [...] the penalty area inside ‘The ball rolls [...] into the penalty area’ I do not see how het strafschopgebied ‘the penalty area’ in (18), for example, can be interpreted as a path (see for more objections Luif 2000: 190, Gehrke 2007: 108). It rather is, in the words of Luif (2000), a spatial domain that is larger than the path of motion. In examples such as (17) and (25), on the other hand, the referent of the non-subject nominal has only a very minimal extension where the subject referent passes. The generalization therefore does not seem to be that the non-subject nominal denotes a path. Rather, the subject referent moves along a path with respect to the referent of the non-subject nominal.

If we reject the generalization that the non-subject nominal of a ParC denotes a path, the difference in acceptability between the PreC in (27) and the ParC in (28), both from Helmantel (2002: 77), has to be explained in a different way. (27) Jan stapt op de kiezelsteen ‘Jan steps on the pebble’ (28) #Jan stapt de kiezelsteen op Jan steps the pebble on Helmantel observes that (28) does not allow interpretations in which Jan is a human being of normal height, and the pebble has the size pebbles normally have. It only allows interpretations in which Jan is very small, or the pebble very large (cf. Schermer-Vermeer 1970: 97). Helmantel accounts for this by referring to her generalization that the nominal that precedes the adposition is interpreted as a path, but this generalization turns out to be problematic. This raises the question how the difference in acceptability between (27) and (28) can be explained. A second problem in Helmantel’s analysis concerns predicate prePPs with path adpositions, of which the prePPs in (23), (24), and (26) are examples: door de tunnel ‘through the tunnel’, door de klapdeuren ‘through the swinging doors’, and over de grens ‘across the border’. Predicative prePPs are characterized by Helmantel in terms of point-like notions: they are said, for example, to give the ‘result location’ (1998: 375) or ‘endpoint’


of the motion (2002: 73), or to “denote some kind of transition point” (2002: 115). These notions make immediate sense in the case of prePPs with point adpositions, which after all are analyzed as denoting point-like locations. It is not clear, however, how such characterizations can be reconciled with Helmantel’s analysis of prePPs with path adpositions as denoting extended locations: how can they denote a point if they denote an extended location? Helmantel does not address this question.

We could, of course, assume that in the case of predicative prePPs with path adpositions, the point is the endpoint of the path that is part of the semantics of the adposition. That would be in line with many semantic analyses of adpositions (see e.g. Bennett 1975, Lakoff 1987, Jackendoff 1983, Cuyckens 1991, Mackenzie 1992),11 and it seems to account for the prePPs in (23), (24), and (26). Additionally, it would illuminate Helmantel’s discussion of (29), from Helmantel (2002: 116, my glosses and translation, MB): (29) *Jan is 30 meter over de brug gefietst Jan is 30 meters over the bridge cycled ‘Jan cycled 30 meters over the bridge’ Helmantel accounts for the ungrammaticality of this example by saying that since predicative prePPs “denote some kind of transition point” (2002: 115), they cannot be combined with a modifier such as 30 meter ‘thirty meters’: “[m]odifying a point by a length modifier does not make sense” (2002: 115).

The problem is, however, that such constructions are in fact possible, cf. the following attested examples: (30) Gerard is een meter of 5 door de lucht gevlogen12 Gerard is a meter or 5 through the air flown ‘Gerard flew about 5 meters through the air’ (31) We zijn een heel stuk door België gereden13 we are a whole bit through Belgium driven ‘We drove a long way through Belgium’

11 Lakoff (1987), for example, states that “[i]t is common for words that have an image schema with a path to also have the corresponding image-schema with a focus on the end point of the path” (1987: 440). Cuyckens (1991) posits a separate sense for this type of usage of Dutch spatial prepositions. Jackendoff (1983) formulates a rule to derive a place expression from a path expression. 12 Http://www.autocross.nl, June 2007. 13 Http://www.dance4life.nl/schools_opvallende_acties_vervolg, June 2007.

40 Chapter 3

(32) Zhen, mr. Tian en ik zijn een heel eind over het strand gelopen14 Zhen, Mr Tian and I are a whole end over the beach walked ‘Zhen, Mr Tian and I walked a long way along the beach’ These constructions feature the auxiliary zijn, which is considered to be diagnostic for predicative prePPs. Yet the prePPs in (30)-(32) do not seem to convey an endpoint, result location, or transition point: they seem to specify paths.

This brings me to a third problem, which is not particular to Helmantel’s analysis. It is a problem for all analyses that somehow relate auxiliary choice in the case of manner of motion verbs to two types of PPs (or the absence of a PP, see Hoekstra 1984, Hoekstra and Mulder 1990, van Hout 1996, Lieber and Baayen 1997, Sorace 2000, cf. Chapter 5 for a detailed discussion). The problem is that actual language use shows that the use of hebben and zijn cross-cuts the two-way PP distinctions that have been proposed. I will give four examples: (33) and (34) with point adpositions, and (35) and (36) with path adpositions. (33) contains hebben, yet the prePP is interpreted as the endpoint of the motion, not as the location where the motion takes place. (33) mijn ventje had van de kant in het water gesprongen15 my guy-DIM had from the side in the water jumped ‘my little one had jumped from the side [of the pool] into the water’ The example is from a mother’s proud story about a successful visit to the swimming pool: her son, the subject referent, has jumped into the water for the first time in his life. While an iterative interpretation is possible for (33), which would be an explanation for the use of hebben, (33) also allows an interpretation in which the subject referent jumps only once: there may have been only one instance of the boy’s brave move. It is this type of interpretation that poses a problem for existing analyses. (34), conversely, contains zijn, yet the prePP is interpreted as the location where the motion event hij is gesprongen ‘he jumped’ takes place.

14 Http://goudtrotter.waarbenjij.nu/, June 2007. 15 Http://tweekids.web-log.nl/twee_kids/index.html, July 2007.


(34) Waarom is hij gesprongen in Amsterdam?16 why is he jumped in Amsterdam ‘Why did he jump in Amsterdam?’ The PreC in bold is a description of how Dutch musician and painter Herman Brood died: he jumped from the roof of the Amsterdam Hilton. A second problematic pair is (35)-(36): they too take the exact opposite auxiliaries from what existing analyses would predict. (35) de juffrouw die met haar paard

the lady who with her horse door een brandende hoepel had gesprongen17 through a burning hoop had jumped ‘the lady who had jumped through a burning hoop with her horse’

(36) Hoe kan het dat de kapers twee uur how can it that the hijackers two hours door Amerikaans luchtruim zijn gevlogen through American airspace are flown

zonder te zijn onderschept door straaljagers van de luchtmacht? 18 without to be intercepted by fighter-jets of the air force ‘How is it possible that the hijackers flew through American airspace for two hours without having been intercepted by air force fighter jets?

(35) contains the auxiliary hebben ‘to have’, which should mean that the prePP is adverbial. Remember that Helmantel characterizes adverbial prePPs with path adpositions as denoting an extended location where the motion takes place. That does not seem correct in the case of (35): a hoop simply does not have the right dimensions for it. In fact, the idea of a result location (or endpoint or transition point) seems to apply here: after the jump the lady is through the hoop, i.e. on the other side of it. Just as (33), (35) allows an iterative reading, but one jump suffices too. In fact, since it is such a dangerous move, a reading in which the woman jumps only once is highly plausible. (36), conversely, contains the auxiliary zijn ‘to be’, which should mean that the prePP is predicative. Helmantel’s point-like characterizations of predicative prePPs do not apply here, however, just as they do not in (30)-

16 Http://www.condoleance.nl/register.php?id=145&start=1960, March 2008. 17 Http://www.dbnl.org/tekst/looy001jaap01, October 2004. 18 ‘Het complot van 11 September [The 9/11 conspiracy]’, Zembla, VARA/NPS, September 10, 2006.

42 Chapter 3

(32). Door het Amerikaanse luchtruim ‘through American airspace’ seems to specify a path.

A final problem concerns the length of the path denoted by postPPs: where must they minimally start and where must they minimally and maximally end? Helmantel’s diagrams represented as Figs. 1 and 2, repeated here, seem to suggest that the paths denoted by postPPs with binnen ‘inside’ and over start somewhere outside the referent of the non-subject nominal.

That cannot, however, be the right generalization for the ParC in (37), in which the referent of hun fietspontje ‘their bicycle ferry’ moves from one side of het kanaal ‘the canal’ to the opposite side. On the basis of what we know about ferries, we assume that the ferry’s path in (37) does not extend beyond the canal’s boundaries in the way Fig. 2 suggests. (37) hun fietspontje [vaart] het kanaal over (jgdapr95)


Nor does the ParC in (38) from Helmantel (2002: 73) require that the man’s climbing starts outside the boundaries of the ladder: Helmantel observes that the man can start to climb from a position on the ladder. (38) De man is de ladder op geklommen

the man is the ladder on climbed ‘The man has climbed up the ladder’

The same can be observed in examples with verder, see (39). Here too the path does not start outside het bos ‘the wood’: the subject referent is already in the wood when his path starts.

x

binnen (x)

Fig. 1

x

over (x)

Fig. 2


(39) Dus liepen ze nog verder het bos in19 so walked they still further the wood in ‘So they walked still further into the wood’

The question where the path can or must end is particularly pressing in the case of point adpositions. Consider, for instance, (40) and the motion events represented in Fig. 3. (40) De tram reed een heuvel op (jnlmar92)

the tram drove a hill on/up ‘The tram drove up a hill’

All four motion events in Fig. 3 are in accordance with Helmantel’s generalizations for postPPs with point adpositions. The paths in Fig. 3 are of varying lengths, which should not be a problem. With point adpositions, according to Helmantel, there is no restriction on the length of the path: it is not fixed, i.e. it may occupy part of what Helmantel calls object x. Furthermore, at the end of each path represented in Fig. 3, the tram is op de heuvel ‘on the hill’, which ties in with the implication that Helmantel formulates for postPPs with point adpositions. Yet, only Figs. 3a and 3b represent possible interpretations of (40); Figs. 3c and 3d do not. Consider also (41) and the motion events represented in Fig. 4. (41) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’

19 Http://members.chello.nl/d.vanegdom/page10.html, December 2005.

Fig. 3. Paths relative to a hill

Fig. 3c Fig. 3d

Fig. 3b Fig. 3a

44 Chapter 3

Here too, all four motion events in Fig. 4 are in accordance with Helmantel’s generalizations for postPPs with point adpositions. The paths have different lengths and share that at the end of the path, the subject referent is in de Leliestraat ‘in Lily Street’. Yet, while Figs. 4a and 4b are readily acceptable as interpretations for (41), Figs 4c and 4d are not, cf. in this respect the following riddle that shows up in various places and guises on the internet:

(42) Q: Hoe ver kan een hond het bos in lopen? how far can a dog the wood in walk ‘How far can a dog walk into the wood?’ A: tot de helft, daarna loopt de hond het bos weer uit20 to the half then walks the dog the wood again out ‘To the middle, after that the dog is walking out of the

wood’

When the subject referents of (41) and (42) have reached the center area of the referent of the non-subject nominal, they are no longer considered to be engaged in the motion events that the constructions express. Interestingly enough, when we provide more context to these examples, longer paths are suddenly possible. If we know, for example, that de Leliestraat ‘Lily Street’ in (41) is a dead-end street, then interpretations such as Figs. 4c and 4d are possible too. Similarly, if het bos ‘the wood’ in (42) is surrounded by a high brick wall with only one opening through which the dog enters, it can walk all the way to the opposite end of the wood. Helmantel’s generalizations for postPPs with point adpositions cannot explain why some motion events are not acceptable as interpretations, nor can they explain why different contexts should have any effect on the

20 Http://www.kantjils.nl/KK/KK_2002_jun.pdf, January 2007.

Fig. 4a Fig. 4b

Fig. 4d Fig. 4c

Fig. 4. Paths relative to a street


acceptability of different lengths of the path relative to the referent of the non-subject nominal. 3.5 Conclusion Helmantel (2002), the most extensive analysis of the semantic differences between ParCs and PreCs, leaves several questions unanswered, which warrants further investigation into the semantics of the two constructions. Section 3.4 identified four problems, the last and first of which concern the semantic analysis of ParCs. They are addressed in Chapter 6, which argues that a more unified analysis is possible than the one that is offered by Helmantel. While she proposes different generalizations depending on (i) the dimensions of the non-subject nominal (those that can ‘function as a path’ and those that cannot) and (ii) the type of adposition (subject referent is P(x) at the end of the motion in the case of point adpositions, but not in the case of path adpositions), Chapter 6 presents a single semantic generalization for Dutch ParCs.

The second and third problems identified in section 3.4, concerning the analysis of PreCs, are taken up in Chapters 4 and 5 respectively. One of the hypotheses of the present study is that the semantic structure of an adposition is the same irrespective of the construction in which it occurs. Chapter 4 therefore presents a case study: an analysis of the semantics of one of the adpositions that can occur in both ParCs and PreCs, namely over. It shares Helmantel’s idea that ‘path’ is an essential aspect of the semantics of adpositions such as over, and explains how the idea of a ‘point’ in some uses of prePPs with over arises in interpretation. Chapter 5 addresses the implications of the observation that hebben and zijn seem to cross-cut the distinction between adverbial and predicative PPs that Helmantel proposes. It argues that while two types of PreCs need to be distinguished, the choice between hebben and zijn does not correlate with the distinction.

Chapter 4 Adpositional semantics: The case of over 4.1 Introduction This chapter presents a cognitive-linguistic analysis of the semantics of Dutch over, which is one of the adpositions that occur in both ParCs and PreCs. A true achievement of cognitive linguistics is that it has shown how seemingly diverse uses of adpositions or other lexical items can plausibly be related to one another by assuming that semantic structure is conceptual in nature. Different spatial configurations in the world can be referred to by means of the same adposition because the human mind perceives a similarity, i.e. categorizes the spatial configurations as the same.

The present study explores the possibility of a monosemous cognitive-linguistic analysis of over, which may seem a contradiction in terms. Most cognitive-linguistic studies analyze the semantics of adpositions as polysemous, i.e. in terms of a network of different, yet related meanings called ‘senses’ (see e.g. Brugman 1983, Lindner 1983, Lakoff 1987, Cuyckens 1991, Dewell 1994, Kreitzer 1997, and Tyler and Evans 2001, 2003). The present study, however, takes a different view, in which linguistic meanings are “taken as general, maintaining their identity in different contexts” (Verhagen 1997: 7).

The semantic structure of Dutch over is proposed to be the single schematic conceptualization described in (1) and represented in Fig. 1, which is considered to function as a constraint on the interpretation process (Fauconnier 1990, 1994, 1997, Verhagen 1997, 2000, van der Leek and Foolen 2006). (1) Over designates a relation between a trajector and a landmark in

which the trajector is related to the landmark by a mental path that follows a surface of the landmark, and from which a force points to the landmark.

Fig. 1. Dutch over

48 Chapter 4

The rectangular in Fig. 1 represents the landmark, the horizontal arrow represents the path, and the vertical arrow represents the force. The trajector is not represented, because each attempt to do so would result in a representation that is too specific.

The chapter provides support for this proposal by focusing on the (spatial) uses of over in PreCs.1 Section 4.2 presents the network of senses proposed for Dutch over by Cuyckens (1991), complemented by a set of senses proposed for English over by Lakoff (1987). Section 4.3 argues that much of the information included in the senses are aspects of interpretations of constructions in which over is used, but not aspects of over’s semantic structure (see also Dewell 1994, Kreitzer 1997, Tyler and Evans 2001, 2003, Van der Gught et al. 2007). Section 4.4 evaluates the objections that have been raised against monosemous analyses against the background of the proposed analysis, and section 4.5 forms the conclusion. 4.2 A lexical network analysis of Dutch over This section presents an overview of Cuyckens’ (1991) analysis of Dutch over, which is complemented by an aspect of Lakoff’s (1987) analysis of English over. Cuyckens, inspired by Lakoff’s analysis, considers the semantics of Dutch over to consist of thirteen related senses that form a lexical network. Ten of these are considered here: the three senses that are not discussed are not part of Standard Dutch as it is spoken in the Netherlands, the language variety that the present study is concerned with.2

Cuyckens’ analysis is a cognitive-linguistic analysis: it assumes that semantic structure is conceptual in nature, and it describes the semantics of

1 The use of over in ParCs is discussed in Chapter 6; its temporal and other metaphorical uses, cf. e.g. (i) and (ii), fall outside the scope of the present study: (i) Greenpeace gaat over twee weken met een schip naar Koeweit (jgdmay91) Greenpeace goes over two weeks with a ship to Kuwait ‘Greenpeace is going with a ship to Kuwait in two weeks’ (ii) Dat zegt de Nederlandse regering over Indonesië (jgdmar92) that says the Dutch government over Indonesia ‘That is what the Dutch government is saying about Indonesia’ 2 The three senses concern uses of over such as (i) and (ii) below, where Standard Dutch as spoken in the Netherlands would use tegenover ‘opposite’ and boven ‘above’ respectively. (i) het restaurant over het museum ‘the restaurant facing/across from the museum’ (ii) Het vliegtuig ongeluk is gebeurd over [...] de Atlantische Oceaan ‘The plane accident happened over the Atlantic’

Adpositional semantics: The case of over 49

the adpostion in terms of ‘relations’ between entities. Adpositions are what Langacker calls ‘relational predications’ (1991b: 75-76):3

a relational predication presupposes a set of entities, and profiles the interconnections among these entities. [... E]very relational predication shows an asymmetry in the prominence accorded the entities that participate in the profiled interconnections: some participant is singled out and construed as the one whose nature or location is being assessed. This participant is called the trajector (tr) and analyzed as the figure within the relational profile. [...] The term landmark (lm) is applied to other salient participants, with respect to which the trajector is situated.

The notion of profiling in the first line of this quotation is an

important one and requires some explanation. Cognitive linguistics assumes that the semantic structure of any language expression consists of a knowledge structure that serves as background, which Langacker calls base (cf. Fillmore’s notion frame, 1982, 1985, and Lakoff’s notion idealized cognitive model, 1987), and some highlighted portion of that background that Langacker calls profile, which is “a substructure elevated to a special level of prominence within the base, namely that substructure which the expression ‘designates’” (Langacker 1991b: 5).4 The semantic structure for elbow, for example, has as its base the conceptualization of a human arm and profiles only a portion of that. Or, as a second example, consider hub, spoke, and rim, which “all invoke as their base the overall configuration of a wheel but contrast semantically by virtue of profiling different portions of it” (Langacker 1991a: 5; see also Fillmore 1982, 1985 for more examples). Returning now to Cuyckens’ analysis, I use the terms ‘trajector’ and ‘landmark’ where he uses (entity) x and (entity) y.5 The first three senses that Cuyckens distinguishes for Dutch over share that they designate a relation between a trajector and a landmark that are not in contact with each other, what Cuyckens calls ‘non-coincidence’, and in which the trajector follows a path ‘above and across’ the landmark. They are illustrated by (2)-(4):6

3 A predication in Langacker’s theory of Cognitive Grammar is “the semantic pole of a linguistic expression (whether a morpheme or more complex)” (1987: 491). 4 I use the terms designation and profile interchangeably, while technically, designation is the “relation within a semantic structure between the base as a whole and some substructure selected as the profile” (Langacker 1987: 488) and profile is the “entity designated by a semantic structure” (Langacker 1987: 491). 5 This is in accordance with Cuyckens’ later work, see e.g. Cuyckens (1994). 6 Examples (2)-(15) are from Cuyckens (1991: 271-279).

50 Chapter 4

(2) Het vliegtuig vloog over de stad ‘The plane flew over the city’ (3) Het vliegtuig vloog over […] de oceaan ‘The airplane flew over […] the ocean’ (4) De Stiekemerd vloog over het doelwit ‘The Stealth flew over the target.’ In sense over1, illustrated by (2), the landmark is a two-dimensional entity which is traversed, from one boundary to the opposite boundary, by the trajector, see Fig. 2a.7 In over2, the landmark is also a two-dimensional entity, but the trajector starts at some point above the landmark, and ends at some other point above the landmark: if the trajector’s path is projected onto the landmark, the projected path stays within the boundaries of the landmark, see Fig. 2b. In over3, finally, the landmark is construed as a ‘dimensionless point’, which is traversed by the trajector, see Fig. 2c.

Cuyckens considers also uses such as (5) and (6) to be instances of over1 and over2 respectively: (5) De brug loopt over de zeeëngte ‘The bridge runs over the strait’ (6) De elektriciteitsleiding loopt over Belgisch grondgebied ‘The power line runs over Belgian territory

(en niet buiten de grenzen van België) (and not outside of the Belgian borders)’

He does not consider them separate senses “because essentially, over denotes the same path, irrespective of whether it is a physical or a mental path” (1991: 280). The trajectors in (5) and (6) are static, 1-dimensional entities that are extended above a two-dimensional surface of a landmark: “the notion of path arises because the conceptualizer mentally scans the [trajector] along its extension” (1991: 276). (5) is an instance of over1 because the path traverses the landmark from one boundary to the opposite

7 Figs. 2-5 are based on Cuyckens’ Fig. 6.9 (1991: 282).

Fig. 2a Fig. 2b Fig. 2c


boundary; (6) is an instance of over2 because the path starts and ends somewhere above the landmark. Three other senses that Cuyckens distinguishes share that they designate a relation between a trajector and a landmark in which the trajector’s path “coincides with the 2DIM surface” of the landmark (1991: 274), i.e. the trajector and the landmark are in contact. They are illustrated by (7)-(9): (7) Hij liep over het Marktplein

‘He walked across the Marketplace’ (8) De bal rolde over de vloer ‘The ball rolled over the floor’ (9) Hij klom […] over de muur

‘He [climbed] […] across the wall’ In over4, illustrated by (7), the trajector moves from one boundary of the landmark to the opposite boundary, see Figure 3a. In over5, illustrated by (8), the trajector moves along a path while remaining within the boundaries of the landmark, see Figure 3b. In over6, illustrated by (9), the landmark is construed as a barrier and the trajector moves from one end of the landmark to the other, see Figure 3c.

Cuyckens’ sense over7, illustrated by (10) and represented in Fig. 4, involves a trajector that “is located at the end of a mental path that starts from the conceptualizer-observer at one end of the [landmark that is construed as a] barrier [...] and that ends at the other end of [it]” (1991: 277). (10) Hij woont over de brug ‘He lives over the bridge’

The nominal in (10) cannot be replaced by just any other nominal: de heuvel ‘the hill’, de bergen ‘the mountains’, de rivier ‘the river’, and de grens ‘the


C

tr

Fig. 4

52 Chapter 4

border’ are possible, but not, for example, de muur ‘the wall’ or het voetbalveld ‘the football field’ (see also Brugman 1983: 24-26, Vandeloise 1991: 424, and Dewell 1994: 358 on this use of English over). The final three senses that Cuyckens distinguishes are related in that the trajector covers the landmark, which involves coincidence between the trajector and a two-dimensional landmark (1991: 281).8 In over8, see Fig. 5a, the static trajector has such dimensions that it covers some surface of the landmark. This surface may be horizontal, as in (11), or vertical, as in (12). (11) Het deken ligt over de tafel ‘The blanket lies over the […] table’ (12) De haarlok hangt over zijn voorhoofd ‘The lock of hair hangs over his forehead’ In over9, see Fig. 5b, the trajector is a ‘multiplex entity’ (Lakoff 1987: 428), i.e. a large number of individual entities that are construed as a ‘continuous region (or mass)’ that covers a surface of the landmark (Cuyckens 1991: 279), see (13). (13) Hij heeft sproeten over heel zijn gezicht

‘He has freckles all over his face’ In over10, finally, see Fig. 5c, the trajector “follows a path that gradually covers” the landmark (Cuyckens 1991: 279): (14) Ze liepen over heel het voetbalveld ‘They walked all over the [football] field’ (15) De vliegen kropen over heel de kast ‘The flies crawled all over the cupboard’

While Cuyckens (1991) observes with respect to over8 that the surface of the landmark may be horizontal or vertical (and judging from his

8 Non-coincidence is possible in covering cases too, a possibility that Cuyckens overlooks, see Geeraerts (1992) and example (20) below.



examples this is possible in the case of over9 and over10 too), he does not posit separate senses to account for such uses. Lakoff (1987), on the other hand, argues in his analysis of English over that covering senses without specifications for the orientation of the landmark cannot account for the different ‘contact restrictions’ that he observes. Since his observations also apply to Dutch, I present his argument here in some detail.

Lakoff observes that in cases in which the trajector is higher than the landmark, there may be contact between the trajector and the landmark, see e.g. (16), or there may be not, see e.g. (17).9 (16) I walked all over the hill (17) Superman flew all over downtown Metropolis Cases in which the trajector is not higher than the landmark, however, require contact between the trajector and the landmark. In (18) and (19), the landmark has a vertical surface, with respect to which the trajector moves: the trajector is not higher than the landmark. While (18) is acceptable, (19) is not. (18) Harry climbed all over the canyon walls (19) *Superman flew all over the canyon walls Lakoff accounts for the difference in acceptability by distinguishing separate covering senses: so-called rotated schemas, in which the trajector is not higher than the landmark, and which specify that the trajector is in contact with the landmark, and so-called unrotated schemas, in which the trajector is higher than the landmark, and which have no such contact specification. The same ‘contact restrictions’ can be observed for Dutch.10 In ‘unrotated’ cases, i.e. in which the trajector is higher than the landmark, the trajector may be in contact with the landmark, see (11) and (14) above, or not, see (20): (20) De spreeuwen zwermden over de heuvels11 ‘The sparrows were swarming over the hills’

9 Examples (16)-(19) are from Lakoff (1987: 429-430). 10 In fact, the contact restrictions apply not only in covering cases, but also in uses in which the trajector moves or is extended along a path, in Dutch and in English. This would mean that more ‘unrotated’ and ‘rotated’ senses need to be distinguished. 11 Http://www.brakkehond.be/68/verbe1.html, February 2008.

54 Chapter 4

In ‘rotated cases’, i.e. in which the trajector is not higher than the landmark, the trajector must be in contact with the landmark, contrast the acceptability of (12), (13), and (15), in which there is contact, with unacceptable (21), for which, just as in its English counterpart, the intended interpretation is that Superman ‘covers’ the vertical surfaces that constitute the landmark: (21) *Superman vloog over de wanden van de kloof *‘Superman flew over the walls of the canyon’ Cuyckens’ covering senses cannot account for these different contact restrictions. 4.3 Dutch over: a single constraint 4.3.1 Introduction This section contrasts Cuyckens’ analysis with an analysis of Dutch over in terms of the semantic structure described in (1) and represented in Fig. 1, repeated here, which is considered to function as a constraint on the interpretation process. (1) Over designates a relation between a trajector and a landmark in


The section argues that many aspects of the senses proposed by Cucykens and Lakoff are not part of the semantics of over, but that they are aspects of the interpretations of the larger constructions in which over occurs (see also Dewell 1994, Kreitzer 1997, Tyler and Evans 2001, 2003, Van der Gught et al. 1997). These aspects are therefore not part of my semantic analysis of Dutch over, which results in a much more schematic characterization. It does not specify: (i) the dimensions of the trajector or the landmark (except that the landmark must have a surface), (ii) whether there is contact between the trajector and the landmark, (iii) whether the path

Fig. 1. Dutch over


crosses or stays within the boundaries of the landmark, (iv) whether the trajector is static or moving, (v) whether the trajector covers the landmark or not, or (vi) whether the path (and by implication the trajector) is higher, lower, or at a similar height as the landmark. Subsections 4.3.3-4.3.8 argue for each of these aspects in turn that they are aspects of the interpretations of larger constructions in which over occurs and should therefore not be included as aspects of over’s semantic structure.

My proposal is in one crucial respect more specific than earlier analyses of over: it includes the notion of a force. This, as subsection 4.3.9 argues, obviates the need to posit separate senses for uses of over in which the trajector is not higher than the landmark.12 Subsection 4.3.2 first of all offers an outline of the theoretical background of the analysis, i.e. the idea of interpretation as constraint satisfaction (Fauconnier 1990, 1994, 1997, Verhagen 1997, 2000). 4.3.2 Linguistic meanings as constraints A basic tenet of cognitive linguistics is that language expressions highly underdetermine the rich interpretations that language users associate with them (e.g. Fauconnier 1994, 1997, Langacker 1987). Linguistic meanings, i.e. the information coded by the language expression itself, do not determine full interpretations; they are ‘clues’ or ‘prompts’ for the language user to construct an interpretation, in the process of which (s)he uses all kinds of other information. Lexical network analyses such as Lakoff (1987) or Cuyckens (1991) have been criticized (see van der Leek 2000, Janssen 2003, Tyler and Evans 2003) for being inconsistent with this basic tenet of cognitive linguistics.

A more precise articulation of how language expressions function as ‘clues’ or ‘prompts’ can be found in Fauconnier (1990, 1994, 1997) and Verhagen (1997, 2000), who define interpretation as a process of constraint satisfaction. The semantic structure of (the words and constructions of) a language expression does not determine a full-fledged interpretation, but constrains the interpretation process, leaving open many possible

12 Van der Gught et al.’s (2007) analysis of English over is very much in the same spirit of the analysis presented here, in that they propose that over has a single meaning: “over invariably expresses that one object is situated in a specific relation to a lower reference point” (2007: 746). They account for the difference between what Lakoff calls ‘unrotated’ and ‘rotated schemas’ “by distinguishing a focal from a disfocal perspective” (2007: 746). What they do not address, however, is the different contact restrictions that Lakoff observes, and I do not see how their analysis can account for that.

56 Chapter 4

interpretations. Linguistic meanings are, in other words, considered to function as constraints, and they are “taken as general, maintaining their identity in different contexts” (Verhagen 1997: 7).

The interpretation process is not only constrained by the language expression itself, but also, at the same time, by the context, defined here in the wide sense of Croft and Cruse (2004): it comprises both linguistic and non-linguistic context, i.e. previous discourse, the immediate linguistic environment, the type of discourse (e.g. genre and register), the physical and social context, and ‘a vast store of remembered experiences and knowledge’ (Croft and Cruse 2004: 102-103). The language user’s task is to come up with an interpretation – using any kind of knowledge at his or her disposal – that satisfies both types of constraints, not by building up an interpretation on the basis of the linguistic constraints first, and then adjusting it on the basis of contextual constraints, but by considering both types of constraints simultaneously (see especially Fauconnier 1990: 391, Verhagen 1997: 28; cf. also Janssen 2007 and Hagoort and van Berkum’s 2007 ‘one-step model’).

The semantics of a word (or of a larger language expression) must be distinguished from its interpretation: “[t]he ‘semantics’ of a language expression is the set of constraints it imposes on cognitive constructions; this is a structural property, which is independent of context” (Fauconnier 1990: 400). The ‘interpretation’ of a language expression is something different: it is the complex conceptualization that is built up within the limits of the constraints imposed by the expression and by the context. 4.3.3 Dimensions of trajector and landmark The constraint that I propose for over specifies that the landmark has a surface, but poses no further requirements on the dimensions of the landmark or the trajector. As Geeraerts’ (1992) points out for uses of over in which a trajector moves across the landmark, i.e. crossing its boundaries, “objects, surfaces, and lines can move over each other in all possible trajector/landmark combinations” (1992: 212). Cuyckens’ senses over1, over3, over4, and over6 account for only a selection of these possibilities.

Rather than concluding that we need to add more senses, let us consider Janssen’s (2003) suggestion with respect to Dutch op. He suggests that it is not the zero-, one-, or two-dimensionality of the landmark (see Cuyckens 1994) that is relevant for the use of op, but the ‘conceivability’ of the landmark as a surface. Similarly, I propose for Dutch over that it is not the dimensions of the trajector or the landmark that are critical, but only that the landmark has a surface which the path in over’s semantics can follow.

The surface may be very minimal, as in (22) or (23).


(22) als de muis over de stip gaat13 when the mouse over the dot goes ‘when the mouse moves over the dot’ (23) Rincon [...] reed alleen over de streep in Andorra (jnljul93) Rincon rode alone over the line in Andorra

‘Rincon rode over the (finish) line in Andorra alone’ De stip ‘the dot’ in (22) is very small, but still it has a surface. If it had not, we would not be able to see it. De streep ‘the line’ in (23) is a finish line that a cyclist crosses. The line does not extend in only one dimension: it also has a certain width. It seems fair to say that mathematical characterizations such as a ‘dimensionless point’ (cf. e.g. Cuyckens over3) are not applicable to such examples. For us to be able to perceive an entity it has to have at least two dimensions,14 and it is such entities that we normally talk about. 4.3.4 Contact The constraint that I propose for Dutch over is neutral with respect to whether the trajector and landmark are in contact. This is an aspect of the interpretation of the larger expression in which over occurs; it is not coded by over itself (for similar proposals for English over, see Dewell 1994, Kreitzer 1997, Tyler and Evans 2001, 2003). Remember that Cuyckens provides (2) as an example of a sense of over that involves no contact between a trajector and a landmark, and (7) as an example of a sense of over in which the trajector and the landmark are in contact. (2) Het vliegtuig vloog over de stad ‘The plane flew over the city’ (7) Hij liep over het Marktplein

‘He walked across the Marketplace’ Distinguishing two senses on the basis of (2) and (7) is unnecessary, however: the no contact/contact difference between (2) and (7) can be explained in terms of the different semantics of the verbs involved. Vliegen

13 Http://www.michelklemann.nl/npzk/site/index.htm, March 2008. 14 Cf. Langacker’s note in his description of how we can perceive a black point against a white background (1987: 108, fn. 6):

‘Point’ is not to be understood in the mathematical sense, which would imply zero extension [...]. Instead it is taken as whatever minimal area functions as a quantum of cognitive processing at a given level of organization.

58 Chapter 4

‘to fly’ takes place through the air, so when a flying plane’s trajectory corresponds to a path that follows a surface of a city, this happens without there being contact between the plane and the city. The combination of the lexical items in (7), on the other hand, results in an interpretation that does involve contact. Kreitzer (1997, see also Tyler and Evans 2001, 2003) argues that contact or non-contact should not be specified for English over by pointing to (24), which does not specify whether there is contact between the man (did he climb over it?) or not (did he jump over it?). (24) The man went over the fence Similar examples can be found for Dutch over: (25) onze zeebadplaatsen werden [...] overstroomd met gasten our seaside-resorts were overflowed with guests die over onze Oostgrens waren gekomen (ot91-10) who over our eastern-border were come

‘Our seaside resorts overflowed with guests who had come across our eastern border’

(26) hij raakte verstrikt in het net he got entangled in the net

dat over het zwembad gespannen was (jgdsep95) that over the swimming-pool spread was

‘He got entangled in the net that was spread over the swimming pool’

Gekomen ‘come’ in (25) does not specify whether the guests crossed the border by plane (no contact) or over land (contact), nor does gespannen ‘spread’ in (26) specify whether the net was (slightly) above the swimming pool (no contact), i.e. suspended in the air, or resting on the surface of the water (contact). 4.3.5 Boundary crossing Cuyckens (1991) posits different senses for Dutch over depending on whether the trajector’s path crosses the boundaries of the landmark, see e.g. (22), or stays well within the boundaries of the landmark, i.e. without ‘touching’ them, see e.g. (27).


(22) als de muis over de stip gaat15 when the mouse over the dot goes ‘when the mouse moves over the dot’ (27) Na haar haastige bezoek aan de rododendron after her hasty visit to the rhododendron vliegt de koningin laag over de grond verder (gp93-2) flies the queen low over the ground further

‘After her hasty visit to the rhododendron the queen flies on low over the ground’

The constraint that I propose specifies that there is a path that follows a surface of the landmark, which accounts straightforwardly for the ‘within-boundaries’ cases. The idea of ‘boundary-crossing’ in a case such as (22), I suggest, arises in interpretation: it is not specified in the semantics of over. In (22), the surface of the landmark has such a small size relative to the trajector that it is likely, i.e. in accordance with our knowledge of such events, that the path of the trajector, a computer mouse, is not limited to the path specified by over de stip ‘over/across the dot’, and so the idea of ‘boundary-crossing’ arises. The relative sizes of the landmarks in these examples, very small in (22) and very large in (27), severely constrain the interpretation possibilities: boundary-crossing in (22), no boundary-crossing in (27). Other examples are more open to both types of interpretations, in which case it is the wider context that imposes constraints that will ‘favour’ one type of interpretation over the other. Cuyckens’ example (3), for instance, is provided to illustrate a within-boundaries sense. Notice, however, that a similar construction in the context of (28) does evoke the idea of boundary-crossing. The sentence that follows mentions the trajector’s arrival, from which we can conclude that the path over the ocean has apparently brought the passenger all the way across.16

15 From http://www.michelklemann.nl/npzk/site/index.htm, March 2008. 16 See also Cuyckens’ example (6): (6) De elektriciteitsleiding loopt over Belgisch grondgebied ‘The power line runs over Belgian territory

(en niet buiten de grenzen van België) (and not outside of the Belgian borders)’

Without the addition in brackets (en niet buiten de grenzen van België ‘and not ouside Belgium’s borders’), both non-boundary-crossing and boundary-crossing interpretations would be possible for (6).

60 Chapter 4

(3) Het vliegtuig vloog over […] de oceaan ‘The airplane flew over […] the ocean’ (28) Een Amerikaans passagier vloog over de oceaan an American passenger flew over the ocean

met explosieven in een koffer. with explosives in a suitcase Bij aankomst werd hij urenlang verhoord. (jnlaug95)

upon arrival was he hours-long questioned ‘An American passenger flew over the ocean with explosives in his suitcase. Upon arrival, he was questioned for hours’

Cuyckens’ example (2), conversely, is meant to illustrate a boundary-crossing sense. The similar construction in (29), however, can quite easily receive an interpretation in which the trajector (hij ‘he’) is in the city when he takes off, so that his path starts somewhere above, i.e. within the boundaries, of the city, and that he comes down again after only a short path. (2) Het vliegtuig vloog over de stad ‘The plane flew over the city’ (29) Elke keer als hij die spuitbussen indrukte

every time when he those spray-cans pressed ging hij omhoog en vloog over de stad17

went he up and flew over the city ‘Every time he pressed those spray cans, he went up and flew over the city’

Cuyckens’ examples (2) and (3) do not, in other words, determine a ‘boundary-crossing’ or ‘within-boundaries’ interpretation; instead, both examples impose the constraint that there is a path following a surface of the landmark. The wider context can narrow down the interpretative possibilities, which is the case in (28): the specification of the passenger’s arrival leads us to conclude that his trajectory crossed the ocean’s boundaries. Again, this is not specified by over, but arises in interpretation. The ‘boundary-crossing’ interpretation is also evoked in constructions with static verbs such as zijn ‘be’, zich bevinden ‘be (located)’, and wonen ‘live’, which are discussed in the following section.

17 Http://www.dvd.nl/forum/archive/index.php/t-12584.html, February 2008.


4.3.6 Static or moving trajector The constraint that I propose for over does not specify whether the trajector is moving or not. It specifies that the trajector is related to the landmark by a mental path, which, I suggest, is what such diverse uses as (30)-(33) have in common:18 (30) Rustig wandelden zij over de Brug van Broederschap en Eenheid

(jnlmar94) quietly strolled they over the Bridge of Brotherhood and Unity ‘They strolled leisurely over the Bridge of Brotherhood and Unity’

(31) toen ik net over de Moerdijkbrug was19 when I just over the Moerdijk-bridge was ‘when I was just across the Moerdijk Bridge’ (32) Over de brug [...] woonde Hartog Goldsteen (mcfeb93ove) ‘Over the bridge lived Hartog Goldsteen’ (33) een klein bruggetje dat over een gracht lag (mcdec93ove)

‘a little bridge that lay across a canal’ Whether the trajector moves along the path as in (30), is located at the end of the path as in (31) and (32), or is an extended entity that ‘constitutes’ the path as in (33) is not specified by over, but is due to the context in which over occurs.

The mental path is of the type that Langacker (2000: 299-301) describes to account for the uses of across in (34)-(36): (34) The child hurried across the busy street. (35) Last night there was a fire across the street. (36) The child is safely across the street. In (34), the trajector (elaborated by the child) is ‘a mover’ that “successively occup[ies] all of the points along a spatial path traversing the static landmark” (elaborated by the street) (Langacker 2000: 299). The trajectors in (35) and (36) do not move: a fire in (35) and the child in (36) occupy

18 My proposal is similar to Dewell’s recent (2007) analysis of English around: “the schematic meaning of around [...] is a scanning pattern in which the conceptualizer’s attention moves in order to focus on the location of the trajector. That scanning pattern is the same whether the trajector is objectively moving or stationary” (2007: 383). 19 Http://www.flitsservice.nl/cm_php/main_repo.php?id=3045, March 2008.

62 Chapter 4

C

tr

lm

T

Fig. 6. One sense of across

“only a single position vis-à-vis the landmark” that is “equivalent to the final position” in (34) (Langacker 2000: 299).

In each of these three cases, though, the conceptualizer (the

speaker/the hearer) mentally traces the same path. In (34) the conceptualizer “necessarily scans mentally along [this] path as an inherent aspect of tracking the subject’s motion” (2000: 300). This is represented in Fig. 6, from Langacker (2000: 300).20 The circles in bold represent the different positions of the trajector with respect to the landmark (represented by the square in bold). The dashed arrows represent the mental contact of the conceptualizer (represented by the circle marked with a C) with the path at different points in processing time (represented by the horizontal arrow marked with a T). While (34) is a case of objective motion (the trajector is actually moving), (35) is a case of subjective motion: “the conceptualizer [...] follows the same path subjectively in locating the trajector with respect to a reference point (R)” (Langacker 2000: 300). This is represented in Fig. 7: the trajector is located at a position that is equivalent to the final position in Fig. 6. The horizontal dashed arrow in bold represents the conceptualizer’s mental scanning.

20 For ease of exposition, I have left out in Figs. 6 and 7 the representations of the maximal scope and immediate scope involved that are present in Langacker’s representation.


What these examples share is that their meanings, i.e. conceptualizations, involve a trajector that is related to a landmark by means of a path that traverses the landmark, i.e. ‘the same mental operations figure’ (Langacker 2000: 299) in each of these examples.21 It is such mental operations that I consider the constraint of over to consists of. The difference between the examples in (34) and (35) is explained by Langacker in terms of different degrees of subjectification, i.e. the extent to which the ‘objective’ situation offers motivation for the conceptualization. (36) presents an intermediate case, for which Fig. 7 can also serve as a representation. The trajector the child is located at the end of the mental path. Just as in (35), the child is not moving, so here too we have a case of subjective motion. What is part of the base, however, is that this location is the result of the trajector’s actual movement along the path, like that of the trajector in (34). While the trajector’s objective motion is profiled in (34), it is unprofiled in (36). (36) is an intermediate case, because the objective situation has a ‘vestige’ of objective motion in it (Langacker 2000: 300). To these examples with across, we can also add (37), which is not discussed in Langacker (2000), but comes from Langacker (1991b: 22). (37) There is a bridge across the river. Just as in (35) and (36), the trajector (the bridge) does not move. Instead, it “simultaneously occupies all the points on a path leading from one side of the primary landmark (the river) to the other” (1991b: 22). Langacker does

21 While across requires that the path goes from one boundary of the landmark to an opposite boundary, Dutch over, I propose, does not. The implications of the difference are addressed below.

C

tr

lm

T

Fig. 7. An extended sense of across

R

64 Chapter 4

not analyze the example any further there, but clearly the same mental path is involved as in the other two senses: the mental path that is necessary to conceptualize the ‘elongated, path-like object’ (1991b: 158) that a bridge is. While Langacker analyzes the uses of across in (34), (35) and (37) in terms of different senses of across, I consider them different types of interpretations, see section 4.4.4 below, i.e. interpretations of across used in different types of contexts. As I stated for Dutch over at the beginning of this section, whether the trajector is static or moving depends on the context in which over is used. What (30)-(33), to which I return in a moment, share is that the trajector is related to the landmark by a mental path. That is what is specified by over.

A difference between across and Dutch over is that across specifies that the path traverses the landmark from one boundary to an opposite boundary, while Dutch over, as I proposed in 4.3.5, does not. This accounts nicely for examples such as (30), which require that the trajector moves along a path that follows a surface of the landmark and allow interpretations in which the path traverses the landmark completely. (30) Rustig wandelden zij over de Brug van Broederschap en Eenheid

(jnlmar94) quietly strolled they over the Bridge of Brotherhood and Unity ‘They leisurely strolled over the Bridge of Brotherhood and Unity’

It also accounts nicely for examples such as (38), with a static extended trajector (see also section 4.3.7 on covering). (38) requires that the trajector een donkerbruine smurrie ‘dark brown sludge’ extends along a path that follows a surface of the landmark, but it also allows interpretations in which the trajector is extended from one of the landmark’s boundaries to an opposite one. (38) […] over de rotsen ligt een donkerbruine smurrie (gp91-4)

‘over the rocks lies dark brown sludge’ That in (33) the boundary-crossing interpretation seems the most natural one is due to our knowledge about this particular combination of trajector and landmark: bridges usually span the entire width of a body of water. (33) een klein bruggetje dat over een gracht lag (mcdec93ove)

‘a little bridge that lay across a canal’


How can we explain, however, that in cases such as (31) with zijn ‘be’, (32) with wonen ‘live’, or (39) with zich bevinden ‘be located’, the trajector is always located at the end of a path that ‘spans’ the entire landmark? (31) toen ik net over de Moerdijkbrug was22 when I just over the Moerdijk-bridge was ‘when I was just across the Moerdijk Bridge’ (32) Over de brug [...] woonde Hartog Goldsteen (mcfeb93ove) ‘Over the bridge lived Hartog Goldsteen’ (39) De olifant […] bevond zich […] over de grens (pb95-04)

the elephant found itself over the border ‘The elephant was located across the border’

The key to the explanation for this lies, in my view, in constructions such as (31), in which the trajector’s static location is the result of his or her (unprofiled) past, actual movement. When the trajector is an entity that is capable of motion, zijn in combination with a prePP with a path adposition such as over evokes this idea of a result. In such cases, the idea of traversal arises by default: the trajector has somehow completed an over path; it is assumed that it was a ‘maximal’ over-path, i.e. a path that follows a surface of the landmark from one boundary to an opposite boundary. If this default situation does not apply, modifiers are used, such as half ‘half’, bijna ‘almost’, nog niet ‘not yet’, etc.: (40) terwijl zijn tege[n]ligger al bijna over de brug was (mcdec95ove) while his oncoming-car already almost over the bridge was ‘while an oncoming car was almost across the bridge’ (41) De bal was al over de doellijn23 ‘The ball was already across the goal line’ The same types of paths, i.e. from one side of the landmark to the opposite side, are involved in (32) and (39), but the motion is purely subjective. It is not the trajector that moves or has moved along the path, but the conceptualizer’s mental scanning. The conceptualizer makes mental contact with the trajector when the conceptualizer’s attention is over de brug ‘over/across the bridge’ or over de grens ‘over/across the border’.

22 Http://www.flitsservice.nl/cm_php/main_repo.php?id=3045, March 2008. 23 Http:// www.homeofthelegends.nl/default.asp?navid=2&nuid=811, March 2008.

66 Chapter 4

4.3.7 Covering This section explains why I do not distinguish one or more separate ‘covering’ senses to account for uses such as (12)-(14). (12) De haarlok hangt over zijn voorhoofd ‘The lock of hair hangs over his forehead’ (13) Hij heeft sproeten over heel zijn gezicht

‘He has freckles all over his face’ (14) Ze liepen over heel het voetbalveld ‘They walked all over the [football] field’ Let us consider (14) first. The idea of ‘covering’ is not due to over, but to the presence of heel ‘entire, whole’. Had heel not been there, this would have been a case of a non-covering sense in Cuyckens’ analysis. Multiplex sproeten ‘freckles’ and de haarlok ‘the lock of hair’, which presumably has quite a bit of extension, can be dealt with in the same way as uses in which an elongated trajector statically extends along the path, except that conceiving of the trajectors in (12) and (13) requires the mental scanning of multiple paths of the ‘over-kind’. To conceive of the freckles in (13), the language user scans multiple paths that follow the surface of zijn gezicht ‘his face’. The information that the conceptualizer scans multiple paths need not be specified in the semantics of over, but is due to the combination of over and the kind of entities involved in (12) and (13), and the information that the motion in (14) takes place with respect to the entire football field. Lakoff distinguishes several covering senses to account for the different contact restrictions. These different restrictions can in fact be observed for non-covering cases of over as well, both in English and in Dutch. The constraint that I propose for Dutch over can account for these restrictions because it includes the information that there is a force pointing from the path to the landmark. This force is the topic of the final section. 4.3.8 A force My constraint for Dutch over does not specify whether the path that follows the surface of the landmark is higher than, lower than, or ‘alongside’ the landmark. Instead, it specifies that there is a force that points from the path to the landmark, by which I mean the following: whatever is located on the path is subject to, or exerts, a force due to which it has a tendency to move


towards the landmark.24 Often the force from the path to the landmark is gravity,25 namely in cases in which the path is higher than the landmark, i.e. further away from the earth’s surface. In (42) and (43) the path specified in the semantics of over corresponds to the trajector’s path of (objective) motion. (42) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrockcanal ‘We are sailing along the Haren-Rüttenbrock canal’ (43) Hij sprong over de toonbank (mcaug94ove) ‘He jumped over the counter’ The participants that move along the path are pulled towards the landmark by gravity. We know, for example, that should the man’s jump in (43) be interrupted somehow while he was in midair above the counter, gravity would pull him down (cf. Freyd et al.’s 1988 study suggesting that even with objects that do not move, human beings are aware of the forces that are working on them). In (38) and (44) too, the force that points from the path to the landmark is gravity, but here the path specified in the semantics of over corresponds to the mental path, or in fact the multiplicity of mental paths, that are necessary to conceive of the trajectors involved. (38) […] over de rotsen ligt een donkerbruine smurrie (gp91-4)

‘over the rocks lies dark brown sludge’ (44) Slingers van heide [...] hingen over de straten (mcaug92ove)

festoons of heather hung over the streets ‘Festoons of heather were hanging across the streets’

24 See Talmy (1988) for the importance of force dynamics for semantic analysis in general. For the importance of force-dynamics for prepositional semantics, see Regier (1996) and Bowerman (1996a), Johnson’s characterization of English in (1987), my semantic analysis of the differences between Dutch aan ‘on’, op ‘on’, and tegen ‘against’ in terms of the different force dynamics involved (Beliën 2002), and Zwarts (forthcoming). 25 As children acquire over well before they learn about gravity, the present analysis could perhaps be improved by stating it in terms of naive physics. The central issue, however, is the force’s effect, i.e. the tendency of whatever is located on the path to move towards the landmark. This is, in other words, irrespective of what is considered to be exerting the force.

68 Chapter 4

The trajectors that are extended along those paths are pulled by gravity towards the landmark. We know, for example, that a suspended festoon that is cut loose from something that it is attached to will drop down to the ground.

Notice that when the force in the semantics of over corresponds to the force of gravity, both contact, see (42) and (39), and non-contact, see (43) and (44), between trajector and landmark are possible, cf. Lakoff’s contact restrictions for ‘unrotated’ cases. The constraint that there is a force that points from the path to the landmark is satisfied either way: gravity pulls any entity towards the landmark if that entity is positioned higher than the landmark, whether that entity is in contact with the landmark or not.

Now let us consider (45) and (46) in which the path is not higher than the landmark. (45) [D]e mieren liepen over het plafond26 ‘The ants were walking over the ceiling’ (46) De waterlanders rollen over haar wangen (jgdfeb94) the waterworks roll over her cheeks

‘Tears are running down her cheeks’ Gravity cannot, in such cases, fulfil the role of the force specified in over’s semantics. In both (45) and (46), the path specified by over corresponds to the path specified by the motion verb. The subject referents that follow these paths are clearly not pulled towards their respective landmarks by gravity. In fact, in (45), gravity is pulling de mieren ‘the ants’ in the opposite direction. We know however that both in (45) and (46) there is another force that is opposes gravity: some kind of (adhesive) force due to which the subject referent tends to move towards its landmark, and it is this force that satisfies the constraint imposed by over. The restriction that there must be contact in ‘rotated cases’ need not be stipulated in terms of different senses in my analysis, but can be considered a ‘side-effect’ of the constraint imposed by over that there must be a force pointing from the path to the landmark. In cases in which the path is not higher than the landmark, gravity cannot fulfil that role, so there must be some other force due to which whatever is located on the path tends to move towards the landmark. Adhesion will do the trick, which necessarily involves contact.27 26 Http://twins.aroundtheglobe.nl/sterke-verhalen/witlof-in-healesville, June 2007. 27 A matter for further research is the following. Melissa Bowerman (p.c.) has drawn my attention to scenarios in which there is a force pointing upward from the path to the landmark, yet over cannot be used. One such scenario involves a helium balloon


Finally, consider examples (35) and (47), which are unlike the other examples in this section in that the paths involved are purely subjective (in the sense of Langacker 2000, Ch. 10): (39) De olifant […] bevond zich […] over de grens (PB95-04)

the elephant found itself over the border ‘The elephant was located across the border’

(47) Hij keek over zijn schouder (GP90-2) ‘He looked over his shoulder’ The subjective paths in (35) and (47) are higher than the landmarks, de grens ‘the border’ in (35) and zijn schouder ‘his shoulder’ in (47). Unlike the other cases we have considered, there is no physical entity located on the path that tends to move towards the landmark due to some force. Still, in these cases too, there is such a force, or actually a whole field of gravitational force, that points from the path to the landmark. 4.4 Discussion 4.4.1 Methodology Lexical network analyses such as Lakoff (1987) or Cuyckens (1991) have been criticized for proposing more senses than necessary or even warranted (Vandeloise 1990, Dewell 1994, Sandra and Rice 1995, Kreitzer 1997, Tyler and Evans 2001, 2003). Sandra and Rice (1995: 92) formulate the problem as follows:

What appears to be a fundamental shortcoming of prepositional network methodology [is] a lack of explicit criteria for distinguishing between usages. Given this vagueness, different linguists are likely to make different distinctions between usage types and to propose different networks for the same preposition.

drifting along a ceiling; another involves a train that runs along just under some kind of magnetic rail (a similar kind of example has been pointed out to me by Dirk Geeraerts, p.c.). What precludes the use of over in such cases seems to have to do with the trajector’s orientation with respect to the landmark: trajector and landmark have to be in some kind of ‘canonical alignment’ (Langacker 1987: 123). A question for further research is therefore whether incorporating a notion along these lines in the constraint description for over would be able to account for the scenarios identified above.

70 Chapter 4

Tyler and Evans (2003: 40), in their analysis of the semantics of English prepositions, state that

[o]ne reason why the number of distinct senses has been exaggerated is that too much importance has been ascribed to the lexical representation, and not enough to the context in which specific interpretations arise. Overemphasizing the information supplied by a particular lexical entry fails to recognize that lexical forms are merely prompts (or in Langacker’s terms ‘access points’) for highly elaborate inferencing and meaning-construction processes.

While they emphasize the role of context, they do not go as far as to propose a monosemous analysis for reasons that are discussed below.

The semantic analysis of over that I have presented has been guided by the hypothesis that semantic structure needs to be general, i.e. that it maintain its identity across contexts (Verhagen 1997: 7). This hypothesis brings along a highly constrained methodology: the linguist’s task is to find what the uses of a word (or a morpheme, or a construction) have in common. The approach is very powerful in that it has led to the idea of a deeper generalization for Dutch over, which includes the notion of a force and an abstraction over cases with a static trajector and those with a moving trajector. The remainder of this chapter evaluates three objections to monosemy against the background of the analysis that I have proposed. 4.4.2 Experimental evidence One problem that Tyler and Evans identify for monosemy relates to experiments such as those conducted by Sandra and Rice (1995) and Rice et al. (1999). The results from these experiments are taken to suggest that “native speakers do associate distinct, conventionalized meanings with a single lexical form” (Tyler and Evans 2003: 37), which would obviously be a problem for monosemous analyses. I think, however, that the findings of those studies can be interpreted differently.

Subjects in Sandra and Rice’s (1995) Experiment 1, for example, were presented with 20 cards, each of which had a sentence on it with the same English preposition. They were asked to group the cards according to “how the preposition was being used in each sentence” (1995: 108). The task makes sense to the subjects, and they consider sentences with spatial uses of the preposition, for example, to be different from sentences with non-spatial uses. From such findings, Sandra and Rice conclude that “a strong monosemy position is untenable” (1995: 89), arguing that if the subjects had


stored in their minds a single meaning for the preposition, they would not consider its uses in some sentences more similar than in other sentences.

Notice, however, that what the subjects are sorting is sentences, i.e. contexts in which a certain preposition occurs. The finding that native speakers consider some contexts to be more similar than others does not, in my view, warrant Sandra and Rice’s conclusion. Another explanation could be that the subjects do have stored in their minds a single meaning for the preposition, but that they can still sort the sentences on the basis of similar or different interpretations of the whole sentences. 4.4.3 Context-independent meanings A second problem for monosemy that Tyler and Evans present is that “some meanings are demonstrably context independent” (2003: 6), i.e. that some uses of a preposition cannot be accounted for on the basis of a single meaning and the context. Let me briefly outline Tyler and Evans’ methodology to elucidate this point. Tyler and Evans’ methodology, which is part of their ‘principled polysemy’ model (Tyler and Evans 2001, 2003), consists of (i) a set of criteria for determining what constitutes the primary sense, or ‘proto-scene’, in a polysemy network and (ii) two criteria for determining whether some use of a preposition counts as a distinct sense. Let us consider the latter two criteria (Tyler and Evans 2003: 42-43) in some detail:

First, for a sense to count as distinct, it must contain additional meaning [...] not apparent in any other senses associated with a particular form, that is, a distinct sense must involve non-spatial meaning or a different configuration between the TR and LM than found in the proto-scene. Second, there must be instances of the sense that are context independent, that is, in which the distinct sense could not be inferred from another sense and the context in which it occurs.

They define the primary sense of English over as a relation in which

the trajector is “higher than but within potential contact” of the landmark (2003: 65), which accounts for examples such as (48), from Tyler and Evans (2003: 40). For (49) and (50), from Tyler and Evans (2003: 43), it is necessary to distinguish a separate sense, which they call the Covering Sense. (48) The humming bird hovered over the flower (49) Joan nailed a board over the hole in the ceiling (50) Joan nailed a board over the hole in the wall

72 Chapter 4

These uses are in conflict with the primary sense, because the trajector a board is not higher than the landmark: in (49) it is lower than the hole in the ceiling and in (50) it is located next to the hole in the wall. Tyler and Evans conclude that “unless we already know that over has a covering/obscuring meaning associated with it, there is no ready contextual means of deriving this meaning in sentences such as” (49) and (50) (2003: 44).

Their Covering Sense is provided as an example of a demonstrably context-independent meaning. Notice, however, that this is not the only possible conclusion on the basis of these data and their two criteria. Another possibility is that the semantic characterization of over that they start out with, i.e. what they consider the primary sense, is not sufficiently general. Since I think that their characterization of the ‘primary sense’ of over is unsatisfactory, I do not share their view that we are dealing with a demonstrably context-independent meaning here. It is my hypothesis that the semantic structure of English over includes a force, just as Dutch over, which can account for cases such as (48)-(50) in a unified way. 4.4.4 Conventionality A third objection to the type of analysis that I have offered can be found in Langacker (1991b), who states that “there is no way to predict precisely which array of extensions and elaborations – out of all those that are conceivable and linguistically plausible – have in fact achieved conventional status” (1991b: 3). Consider, for example, Lindner (1983: 70-71) who claims with regard to English out that

it would certainly oversimplify matters to list only one configuration [...] as the one and only profile of OUT. We need to represent in our account of English the full range of conventional uses of OUT as well as the unity of these uses.

I agree that a semantic analysis of a linguistic item needs to be based

on the full range of conventional uses of that item. Those conventional uses have to be taught, for example, to second language users. I do not agree, however, that those conventional uses are senses, i.e. meanings, of the linguistic item. Instead, they are abstractions over contexts in which the item is used, i.e. abstractions over interpretations, see also Van der Gught et al. (2007: 748) on Tyler and Evans’ senses for English over, such as the Covering Sense, the On-the-other-side-of Sense, and the Above-and-beyond Sense:


What is left unaccounted for in defining the meaning (i.e., enumerating the senses) of over in this way is the fact that the preposition does not mean what the paraphrases suggest as an isolated word. Rather, the suggested paraphrases render the readings of over as part of utterances, i.e. they are concerned with readings in context, in particular with the way over is conventionally interpreted in combination with other words in phrases and clauses.

4.5 Conclusion The analysis of over presented here agrees with Helmantel (2002) that there is a path in the semantic structure of over. The analysis differs from Helmantel’s in that prePPs with over are not considered to denote an extended location; instead, they designate a relation in which the trajector is related to the landmark by means of a path that follows a surface of the landmark and from which a force points to the landmark. The chapter has argued that this generalization obviates the need to account for the semantics of over in terms of a network of distinct, but related senses.

The chapter has further argued that the idea of a point, i.e. at the end of a path that follows the entire surface of the landmark, arises in certain contexts, e.g. when the landmark is very small, or in constructions with zijn ‘be’, wonen ‘live’, or zich bevinden ‘be (located)’. In Helmantel’s analysis, this type of interpretation is ascribed to the predicative function of the prePP, for which the auxiliary zijn is taken as diagnostic. Chapter 3 showed, however, that zijn is also found in PreCs in which the idea of a point does not arise, and that hebben can be found in PreCs in which the idea of a point does arise. It is this problem that is the focus of Chapter 5.

Chapter 5 Prepositional phrases and auxiliary choice 5.1 Introduction The present study hypothesizes that the key to the grammatical constituency of ParCs lies in an analysis of the semantic differences between ParCs and PreCs. Before such an analysis can be presented, it is necessary to address the question whether there are in fact two types of PreCs, i.e. along the lines of Helmantel’s (2002) distinction between constructions with predicative PPs and constructions with adverbial PPs, cf. also Hoekstra (1984), Hoekstra and Mulder (1990), and van Hout (1996). These studies consider the perfect auxiliaries hebben ‘have’ and zijn ‘be’ to be diagnostic for different functions of the PPs. Chapter 3, however, has presented data that are problematic for such claims. The present chapter argues, drawing on the semantic analysis of over presented in Chapter 4, that it is necessary to distinguish two types of PreCs, but that hebben and zijn cross-cut the distinction.

Section 5.2 briefly recapitulates Helmantel’s distinction between predicative and adverbial PPs, and then goes on to discuss a number of other analyses that somehow relate auxiliary choice in the case of manner of motion verbs to two types of PPs (or the absence of a PP). Section 5.3 shows that the attested data presented in Chapter 3, complemented by some additional data, are not only problematic for Helmantel’s analysis, but for the other analyses too, which raises the question of whether auxiliary choice can at all be related to two types of PPs. Section 5.4 investigates this issue from the perspective of Cognitive Grammar and concludes that there are two different ways in which PPs in PreCs semantically integrate with the verb of motion, i.e. there are two PreCs, but that auxiliary choice cannot be related to the two modes of integration. Section 5.5, finally, argues that hebben and zijn construe a motion event in different ways, i.e. as a type of act in the case of hebben and as a change of location in the case of zijn (cf. Geerts et al. 1984, Honselaar 1987, Shannon 1990, and Haeseryn et al. 1997). 5.2 Auxiliary choice and PPs As pointed out in Chapter 3, Helmantel (2002) distinguishes between adverbial and predicative PPs. Adverbial PPs specify the location where the motion takes place, and they occur with hebben. Predicative PPs specify the

76 Chapter 5

‘endpoint’ of the motion (2002: 73; cf. ‘result location’, Helmantel 1998: 375; or ‘transition point’, Helmantel 2002: 115), and they occur with zijn.1

According to this analysis, the prePPs in (1) and (2) have an adverbial function: they occur with hebben. The prePP in (1), which contains the point adposition in ‘in’, denotes a ‘point in space’ where the motion takes place. The prePP in (2), which contains the path adposition door ‘through’, denotes an ‘extended location’ where the motion takes place. (1) Zaterdag hebben de Veenruiters Saturday have the Peat-riders in de Pasmanmanege in Steenwijk gereden (mcmar93spo) in the Pasman-arena in Steenwijk ridden ‘Saturday the Peat Riders rode in the Pasman Arena in Steenwijk’ (2) met de lippen stijf op elkaar had zij

with the lips tightly on each-other had she door het donkere bovenhuis gelopen (de Winter 1989: 198) through the dark upstairs-apartment walked ‘She had walked through the dark upstairs apartment with her lips tightly pressed together’

The prePPs in (3) and (4), conversely, must be analyzed as having a

predicative function: they occur with zijn. Chapter 3 observed that point-like notions such as endpoint, result location, or transition point make immediate sense in cases with point adpositions, such as (3): the motion ends at the point-like location in de sloot ‘in the ditch’. (3) De automobilist A.D. uit Hoensbroek is vrijdag [...] the driver A.D. from Hoensbroek is Friday naast Rijksweg 32 in de sloot gereden (mcaug92spo) alongside highway 32 in the ditch driven

‘Driver A.D. from Hoensbroek drove into the ditch alongside Highway 32 last Friday’

(4) je kon het [meer] pas zien you could it [the lake] only see

1 Helmantel (2002: 18) notes that not all predicative prePPs occur with zijn, only those in ‘ergative’ constructions, i.e. what I call PreCs. In a ‘non-ergative’ construction such as (i), from Helmantel (2002: 18), hebben is used. These constructions fall outside the scope of the present study. (i) dat Jan het boek op de plank heeft gezet that Jan the book on the shelf has put ‘that Jan has put the book on the shelf’

Prepositional phrases and auxiliary choice 77

als je door Bracciano was gereden (de Winter 1989: 377) when you through Bracciano was driven

‘you could only see it [the lake] when you had driven through Bracciano’

Cases with path adpositions, such as (4), are more problematic in this respect. Chapter 3 raised the question how they can denote a point, when they are also considered to denote ‘extended locations’. This issue is not addressed by Helmantel (2002). One way to understand it, though, would be that the point is the subject referent’s location when (s)he is door Bracchiano ‘through Bracchiano’. Extending the analysis of over presented in Chapter 4 to door ‘through’, we could say that door designates a relation in which a trajector is related by a path that is contained in the landmark. A predicative prePP with door could then be a context in which the idea that the trajector follows a maximal path, i.e. a path that the landmark maximally allows: from one end of the landmark to the other end of the landmark.

Helmantel (2002) is not the only analysis that relates auxiliary choice to two types of PPs. Van Hout (1996), for example, considers hebben and zijn to be diagnostic for ‘unergativity’ versus ‘unaccusativity’ (cf. Perlmutter 1978, Burzio 1986, Levin and Rappaport Hovav 1995). She states, along the lines of Zaenen (1993), that Dutch verbs of manner of motion may take just one argument, i.e. the participant that moves, in which case (i) they are atelic, (ii) they occur in an unergative syntactic frame, and (iii) they select hebben. Such verbs may also take an extra ‘goal’ argument “toward which the motion is directed” (van Hout 1996: 63), in which case (i) they are telic, (ii) they occur in an unaccusative syntactic frame, and (iii) they select zijn.

If we apply van Hout’s analysis to (1)-(4), we can say that the prePPs in the unergative frames in (1) and (2) are not arguments, but modifiers of the atelic motion event rijden ‘ride’, while the prePPs in the unaccusative frames in (3) and (4) are goal arguments, which make the motion event rijden ‘drive’ telic. Notice that here too, we could raise the question how door Bracciano ‘through Bracciano’ is to be understood as a goal argument ‘toward which the motion is directed’.

Telicity is similarly crucial in the analysis of Sorace (2000), who claims that Dutch manner of motion verbs ‘normally select’ hebben, but “switch from hebben to zijn when they are embedded in a predicate that has been telicized by a directional phrase” (2000: 875). Notice that while van Hout characterizes the PPs found with zijn as ‘goal’ arguments, Sorace uses the term ‘directional phrase’. This is problematic in light of recent research (Helmantel 1998, 2002, Luif 2000, Beliën 2002, and Gehrke 2007) that

78 Chapter 5

suggests that prePPs with in ‘in’ or op ‘on’ should not be analyzed as directional, but as purely locative, ‘even’ in constructions such as (2). The directionality of (2) is due not to the semantic structure of the prePP, but to the larger construction in which it occurs. The attraction of such proposals is of course that they offer a more encompassing generalization: prepositions such as in or op do not have a locative and a directional sense, but are purely locative; any directionality is due to the contexts in which they occur. I conclude therefore that ‘directional phrase’ does not appropriately characterize PPs that occur with verbs of manner of motion taking zijn.

Lieber and Baayen (1997), finally, also propose a generalization that features the term ‘directional phrase’: “unergative verbs like lopen ‘walk’ which normally take the auxiliary hebben can be used with zijn just in case a directional phrase is added” (1997: 807). The same objections could be raised here, but Lieber and Baayen modify their claim in such a way that it becomes somewhat more precise: “it is not the mere presence of a directional phrase but its actual interpretation as adding an inferable eventual position that is crucial for the change in auxiliary selection” (1997: 808).

To see what they mean by ‘inferable eventual position’, consider the following quote (from Lieber and Baayen 1997: 795-796):

so-called ‘manner of motion verbs’ [...] do not allow any inference as to the eventual position of their highest argument. For example, the verb lopen ‘walk’ by itself implies nothing about the eventual position of its highest argument, the theme; the theme may end up anywhere, including where it started.

Their example of a directional phrase that is not interpreted as adding an inferable eventual position, and therefore takes hebben, is (5), from Lieber and Baayen (1997: 808). They state that it “is acceptable only on the interpretation that the event of walking described was one of wandering or pacing aimlessly around in the room” (1997: 808): (5) Hij heeft door de kamer gelopen he has through the room walked ‘He walked through the room’

The analyses discussed in this section can be summarized as follows:

zijn is found with prePPs that are interpreted as the endpoint of a telic motion event, while hebben is found with prePPs that give the point-like or extended location where an atelic motion event takes place.


5.3 Problematic attested data This section presents data that are problematic for the analyses discussed in section 5.2: they take the exact opposite auxiliary from what the analyses would predict. Some of the examples were already discussed in Chapter 3; others are new.

Consider first of all the PreCs in (6)-(9). They describe motion events that may have occurred only once: we could add, for example, the modifier één keer ‘once’. These non-iterative interpretations present a problem for the analyses discussed in the previous section. They contain hebben, yet the motion events are not atelic, nor do the prePPs give the (point-like or extended) location of the motion event. Instead, they are telic: they have clear endpoints, namely when the girl in (6) is in the swimming pool, the little boy in (7) is in the water, the lady in (8) is through the burning hoop, and the man in (9) is over, i.e. on the other side of, (a bar that is placed at a height of) 2m10. (6) Het meisje had in het zwembad gedoken2 the girl had in the swimming-pool dived ‘The girl had dived into the swimming pool’ (7) mijn ventje had van de kant in het water gesprongen3 my guy-DIM had from the side in the water jumped ‘my little one had jumped from the side [of the pool] into the water’ (8) de juffrouw die met haar paard


(9) dat hij over 2m10 had gesprongen5 that he over 2m10 had jumped ‘that he had vaulted over 2m10’

(10), conversely, features zijn, yet the prePP is not interpreted as the

endpoint of the motion. It is understood as the location where the motion event hij is gesprongen ‘he jumped’ takes place. That event is telic, but it is not ‘telicized’ by the prePP. 2 Http://zwembadgids.nl/zwembadnieuws.asp, July 2007. 3 Http://tweekids.web-log.nl/twee_kids/index.html, July 2007. 4 Http://www.dbnl.org/tekst/looy001jaap01, October 2004. 5 Http://www.kvac.be/clubblad/490204.htm, October 2004.

80 Chapter 5

(10) Waarom is hij gesprongen in Amsterdam?6 why is he jumped in Amsterdam ‘Why did he jump in Amsterdam?’ The PreC in (11) also contains zijn, yet expresses a motion event that is not telic, witness the modifier twee uur ‘for two hours’. Nor is the PP door Amerikaans luchtruim ‘through American airspace’ understood as giving the endpoint of the motion or an ‘inferable eventual position’. Instead it seems to specify a path. (11) Hoe kan het dat de kapers twee uur how can it that the hijackers two hours door Amerikaans luchtruim zijn gevlogen through American airspace are flown


The PreC in (11) as such allows interpretations in which the hijackers are flying criss-cross within the boundaries of American airspace, possibly ending up at the same position as where they started from. Because the 9/11 context of (11) possibly makes it difficult to think of such interpretations (our knowledge of the context imposes constraints that do not allow such interpretations), let us consider the PreC in (12), which illustrates the same point: (12) Een 25-jarige man uit Maasdijk is dinsdagmiddag in zijn blootje

a 25-year-old man from Maasdijk is Tuesday-afternoon in his nudity door het Spijkerkwartier in Arnhem gerend, op zoek naar een through the Spijker-quarter in Arnhem run, in search for a prostituée die […] zijn portefeuille […] had gestolen (mcjun95ove) a prostitute who […] his wallet […] had stolen ‘Tuesday afternoon a 25-year-old man from Maasdijk ran naked through the Spijkerkwartier in Arnhem, looking for a prostitute who had stolen his wallet’

6 Http://www.condoleance.nl/register.php?id=145&start=1960, March 2008. 7 ‘Het complot van 11 September [The 9/11 conspiracy]’, Zembla, VARA/NPS, September 10, 2006.


We can imagine the man running within the boundaries of het Spijkerkwartier in Arnhem, a neighborhood in the city of Arnhem, going here and there, trying to find the prostitute, possibly ending up at the position he started from. The sentence that follows after (12), namely (13), makes it clear that the path stays within the boundaries of the landmark; the chase only lasts for a few streets: (13) De naakte Maasdijker achtervolgde haar door enkele straten, ‘The naked man from Maasdijk chased her through a few streets,’ maar raakte de vrouw kwijt (mcjun95ove) ‘but then lost the woman.’ A final example of a PreC with zijn in which the PP seems to specify a path within the boundaries of the landmark, rather than an endpoint, is (14). (14) Maar of [... koningen, prinsen en prinsessen en hun gevolg] ooit

But whether kings, princes, and princesses and their retinue ever over de weg van Ruinen naar Echten zijn gereden is niet bekend (mcdec95ove) over the road from Ruinen to Echten are ridden is not known ‘But whether kings, princes, and princesses and their retinue ever rode along the road from Ruinen to Echten is not known.’

While the construction in (14) allows interpretations in which the referents of the subject move along a maximal path, i.e. one that follows the surface of the landmark from one end to the opposite end, this is not required. The subject referents may also move along a ‘partial’ path, i.e. one that stays (well) within the boundaries of the landmark.

Support for the idea that PPs such as those in (11)-(14) specify paths rather than points comes from data such as (15)-(17), which contradict Helmantel’s (2002: 115-116) intuition that ‘length modifiers’ do not occur with PPs in constructions with zijn. (15) Gerard is een meter of 5 door de lucht gevlogen8 Gerard is a meter or 5 through the air flown ‘Gerard flew about 5 meters through the air’

8 Http://www.autocross.nl, June 2007.

82 Chapter 5

(16) We zijn een heel stuk door België gereden9 we are a whole bit through Belgium driven ‘We drove a long way through Belgium’ (17) Zhen, mr. Tian en ik zijn een heel eind over het strand gelopen10 Zhen, Mr Tian and I are a whole end over the beach walked ‘Zhen, Mr Tian and I walked a long way along the beach’ The PPs in (15)-(17) specify paths, not points, and the length of each path is given by the phrases in bold. In conclusion, the examples presented in this section take the opposite auxiliary from the one expected on the basis of the accounts outlined in section 5.2, which raises the question whether auxiliary choice can in fact be related to two types of prePPs. 5.4 PreCs: a cognitive-grammar perspective 5.4.1 Complex expressions in Cognitive Grammar This section investigates, from a cognitive-grammar perspective (Langacker 1987, 1991a, 2000), whether it is necessary to distinguish two types of PreCs. Cognitive Grammar makes the radical claim that the grammar of a language consists of just three types of basic structures: semantic structure, phonological structure, and symbolic structure, which is a pairing of a semantic and a phonological structure. It does not, in other words, distinguish syntactic structure as a separate level of organization.

Morphemes, words, phrases, clauses, and sentences are all symbolic structures with different degrees of complexity. The Dutch morpheme (and word) over ‘over’, for example, is a simple symbolic structure that consists of the phonological structure [over] and a semantic structure, see my proposal for the semantic structure of over in Chapter 4. The prePP over het strand ‘over the beach’, see (17), is a complex symbolic structure in that it consists of components that are symbolic structures themselves: over ‘over’, het ‘the’, and strand ‘beach’.

Such a complex symbolic structure is called a grammatical construction.11 A grammatical construction consists of “a set of component structures, their mode of integration, and the composite structure resulting

9 Http://www.dance4life.nl/schools_opvallende_acties_vervolg/5276, June 2007. 10 Http://goudtrotter.waarbenjij.nu, June 2007. 11 Note that in Cognitive Grammar, the term ‘grammatical construction’ can be used for specific structures (e.g. in de car) and schematic patterns abstracted from such specific structures (also known as ‘constructional schemas’, see below).


from the integration” (Langacker 1987: 489). The English grammatical construction in the car, for example, can be represented as Fig. 1. It consists of the composite structure in the car, represented by the box at the top. The composite structure is a symbolic structure that consists of a phonological structure, represented as [in the car], and a semantic structure, represented pictorially: it designates a relation between a trajector and a car in which the trajector is contained in the car.

Fig. 1. A representation of the grammatical construction in the car

The construction consists of two component structures: the symbolic

structure in, represented by the lower box on the left, and the symbolic structure the car, represented by the lower box on the right.12 The component structures can integrate to form the more complex composite structure “by virtue of having certain substructures in common” (Langacker 1987: 278). Integration takes place at the phonological pole and at the semantic pole (cf. Langacker characterization of the English prepositional-phrase

12 The car is itself a grammatical construction: it consists of the two component structures the and car. This is ignored in Fig. 1 and my discussion of it.

the car

in the car

lm

tr

in

tr

lm

84 Chapter 5

construction, 1987: 290, 1991a: 153). At the phonological pole, the structures [in] and [the car] are integrated in such a way that the phonological structure [the car] follows the phonological structure [in] directly. The composite phonological structure [in the car] can be found at the top, i.e. the phonological pole of the composite symbolic structure. Notice that a dotted line connects [in] in the composite structure with [in] in the component structure to the left, and another connects [the car] in the composite structure with [the car] in the component structure to the right. These lines are meant to represent correspondences, i.e. positive judgments of identity. The semantic poles of in and the car are integrated in the following way. The lower part of the box on the left represents the semantic structure of in: a relation between a trajector and a landmark in which the trajector is contained in the landmark. The landmark of in is ‘placed in correspondence with’ the conceptualized entity that constitutes the semantic structure of car, and these two structures (the landmark of in and the car) are “construed as designating the same conceived entity” (Langacker 1987: 94). While in on its own symbolizes a containment relation between a trajector and a schematic landmark, the grammatical construction in the car is more specific in that it symbolizes a relation between a trajector and a car, in which the trajector is contained in the car.13 Here too, the dotted lines represent correspondences: the landmark of in and the semantic structure of car, for example, are linked by a dotted line, just as they are linked to the car that functions as landmark in the semantic structure of the composite structure. The component structure in is called the head, or profile determinant, of this construction, because the construction as a whole inherits in’s relational profile: the component structure in profiles a relation, i.e. between a trajector and a landmark, and the composite structure in the car does so too.

Cognitive Grammar assumes that the knowledge of a language is usage-based, which means, among other things, that a child acquires specific expressions first, from which (s)he abstracts generalizations (see also Tomasello 1999). On the basis of such specific grammatical constructions as in the car, in bed, in your bottle, in the water, etc., for example, the child will start to see similarities across these expressions. These similarities are represented in Fig. 2, which is a schematic grammatical construction, or a constructional schema. Constructional schemas can be used to create and understand expressions that one has not heard before.

13 Chapter 6 notes that only partial containment is required.


Fig. 2 represents the following similarities across specific

grammatical constructions such as in the car, in bed, in your bottle, and in the water. First of all, they all consist of the symbolic structure in, which consists of the phonological structure [in] and a semantic structure that designates a relation between a trajector and a landmark in which the trajector is contained in the landmark. Secondly, the phonological structure [in] is followed by a phonological structure that forms a symbolic structure with a semantic structure that designates a thing.14 This thing, thirdly, elaborates the landmark in the semantic structure of in: the two structures,

14 Thing is used here in the cognitive-grammar sense of a ‘region in some domain’, cf. Langacker (1987: chapter 5). Word classes are notionally defined in cognitive grammar. Adjectives, adverbs and adpositions are characterized as profiling ‘atemporal relations’. Verbs are characterized as profiling ‘processes’, which are temporal relations: relations that are viewed as evolving through time. What nouns share is that they profile a ‘region in some domain’, where a ‘domain’ is any “coherent area of conceptualization relative to which semantic units may be characterized” (Langacker 1987: 488).

…

in …

lm

in

lm

Fig. 2. A representation of the constructional schema in NP

tr

tr

86 Chapter 5

the thing and the landmark, are “construed as designating the same conceived entity” (Langacker 1987: 94). This makes the semantic structure of the phrase as a whole more specific than that of in: the phrase as a whole designates a relation in which the trajector is contained in an elaborated landmark. 5.4.2 Two types of PreCs Cognitive Grammar offers, as I see it, two possibilities for a two-way distinction of prePPs to which auxiliary choice could be related. One possibility is that two types of prePPs are distinguished on the basis of their different semantic structures. This, however, does not seem to be the case. As was pointed out before, an attractive aspect of Helmantel’s (2002) analysis is that she considers the semantic structures of the prePPs as such to be the same, regardless of the syntactic configuration in which they occur. The idea is in line with one of the hypotheses of this dissertation that “linguistic meanings must be taken as general, maintaining their identity across contexts” (Verhagen 1997: 7), see Chapter 4. A second possibility is that we distinguish two types of prePPs on the basis of how they are integrated with other elements in the grammatical construction. Consider Langacker’s (1987: 308) examples (18) and (19). (18) I put the sweater in a box. (19) We chased squirrels in the park. In the box in (18) can be integrated semantically with the verb put because it is construed as corresponding to a substructure of the verb’s semantics: in the box ‘elaborates’, i.e. corresponds to and specifies in further detail, the final destination that is part of the semantic structure of put (Langacker 1987: 308). We can call this PP a complement, which is defined in Cognitive Grammar as a ‘component structure that elaborates a salient substructure of the head’ (Langacker 2000: 21). The head, i.e. the profile determinant, is the verb in (18): the construction as a whole inherits the verb’s processual profile. In the park in (19) is integrated with the verb chased in a different way: while the PP in (18) elaborates a substructure of the verb, in (19) the verb elaborates a substructure of the PP: the process designated by we chased squirrels elaborates the trajector of in the park. We can call this PP a modifier, which in Cognitive Grammar is defined as a ‘component structure a salient substructure of which is elaborated by the head’ (Langacker 2000: 21).


These characterizations of the integration of a PP with a verb also generalize to Dutch PreCs such as (1) and (3), repeated here: (1) Zaterdag hebben de Veenruiters Saturday have the Peat-riders in de Pasmanmanege in Steenwijk gereden (mcmar93spo) in the Pasman-arena in Steenwijk ridden ‘Saturday the Peat Riders rode in the Pasman Arena in Steenwijk’ (3) De automobilist A.D. uit Hoensbroek is vrijdag [...] the driver A.D. from Hoensbroek is Friday naast Rijksweg 32 in de sloot gereden (mcaug92spo) alongside highway 32 in the ditch driven


In de Pasmanmanage in Steenwijk ‘in the Pasman Arena in Steenwijk’ in (1) is a modifier: a salient substructure of it, namely its trajector, is elaborated by the process designated by the head rijden ‘ride’. In de sloot ‘in the ditch’ in (3), on the other hand, is a complement: it elaborates a substructure of the semantics of the verb rijden ‘drive’, which requires a construal of the path in the semantics of the verb as having an endpoint: it is this endpoint that is elaborated by in de sloot ‘in the ditch’.15 So, for point adpositions such as in, we must distinguish two types of prePPs on the basis of their different ‘modes of integration’ (Langacker 1987: 277) and conclude that a distinction between two types of PreCs is warranted. I will refer to constructions such as (1) as Modifier PreCs and to constructions such as (3) as Complement PreCs.

The problematic PreCs with point adpositions (6), (7), and (10) can now be analyzed as follows. (6) and (7) are Complement PreCs: the prePP elaborates the endpoint of the motion profiled by the head.

15 Rijden ‘to ride, drive’ is different from put in that put “implies a final destination for its primary landmark, so in a box [in (18)] elaborates a substructure already prominently introduced by put” (Langacker 1987: 308), while a final destination is not prominently introduced by rijden. That is why I propose that the path that is inherent in the semantics of rijden is construed in (3) as having an endpoint, cf. Langacker (1987: 305): “To find within the dependent predication a substructure that is fully schematic for the autonomous component, it may [...] be necessary to carry out a focal adjustment of some kind”, i.e. some type of construal. The substructure that is the result of the construal, i.e. the endpoint of the path, can be elaborated by in de sloot ‘in the ditch’.

88 Chapter 5

(6) Het meisje had in het zwembad gedoken16 the girl had in the swimming-pool dived ‘The girl had dived into the swimming pool’ (7) mijn ventje had van de kant in het water gesprongen17 my guy-DIM had from the side in the water jumped ‘my little one had jumped from the side [of the pool] into the water’ (10) is a Modifier PreC: the motion event profiled by the head elaborates the trajector of the prePP.18 (10) Waarom is hij gesprongen in Amsterdam?19 why is he jumped in Amsterdam ‘Why did he jump in Amsterdam?’ Let us now turn to PreCs with path adpositions, such as (2) and (4) with door ‘through’, repeated here. Unlike previous analyses, I do not consider these examples to involve different types of prePPs. (2) met de lippen stijf op elkaar had zij





My analysis takes as its starting point Langacker’s (1987: 305) characterization of example (20), in which the PP through the tunnel is analyzed as a complement, because it elaborates a substructure of the head.

16 Http://zwembadgids.nl/zwembadnieuws.asp, July 2007. 17 Http://tweekids.web-log.nl/twee_kids/index.html, July 2007. 18 Notice that the prePP occurs here in ‘extraposed’ position, which is generally considered to be a diagnostic for modifier (or adverbial) prePPs. A matter for further research is how such attested ‘extraposition’ data tie in with the analysis presented here. 19 Http://www.condoleance.nl/register.php?id=145&start=1960, March 2008.


(20) The little girl crawled through a tunnel Langacker (1987: 305) writes:

Crawl implies that its trajector follows a spatial path but specifies nothing very precise about its nature. This path functions as e[laboration]-site in the valence relation [between the verb and its prepositional-phrase complement], being elaborated by the more detailed path-specification through a tunnel.

In my view, the modes of integration of the prePPs in (2) and (4) are the same as in (20): door het donkere bovenhuis ‘through the dark upstairs apartment’ and door Bracciano ‘through Bracciano’ specify paths that elaborate the paths in the semantics of the motion verbs gelopen ‘walked’ and gereden ‘driven’. These prePPs each elaborate, to put it differently, a substructure of the head and can therefore be called complements. The idea of a ‘goal’ or ‘transition point’ that is present in (4), and also in (7) and (8), repeated here, is not, in my view, part of the semantic structure of the prePPs as such, nor is it due to a particular mode of integration. Instead, it arises in interpretation. (7) de juffrouw die met haar paard



The constraint imposed by a prePP with door ‘through’ or over ‘over’ is that the trajector is related to the landmark by means of a path. This path is not specified for its length, shape, or direction; the constraint only specifies that the path, in the case of door, is contained in the landmark, or, in the case of over, follows a surface of the landmark with a force that is directed from the path to the landmark (see Chapter 4).

When such prePPs elaborate the path inherent in the semantics of a verb of motion, there are all kinds of interpretations that can satisfy the constraints imposed by the construction. In a construction such as (21), we

20 Http://www.dbnl.org/tekst/looy001jaap01, October 2004. 21 Http://www.kvac.be/clubblad/490204.htm, October 2004.

90 Chapter 5

think of a path that starts and ends somewhere within the boundaries of the landmark, due to its vastness.22 (21) Gerard vloog door de lucht ‘Gerard flew through the air’ Such paths can be modified by means of what Helmantel calls ‘length modifiers’, as we saw in (15)-(17). In constructions such as (7) and (8), on the other hand, we think of a path that starts on one side of the landmark and ends on the other side of the landmark. Just as Tyler and Evans (2003: 70-71) argue for The cat jumped over the wall, so I argue that what is profiled by the prePP door een brandende hoepel ‘through a burning hoop’, for example, is only a small portion of the path that is involved in the interpretation of the PreC as a whole.23 This prePP profiles only the very short path that is ‘contained’ in the landmark, the hoop. This very short path elaborates the path involved in the semantics of springen ‘to jump’, which “codes self-propelled motion using the ground to push off from” (Tyler and Evans’ description of the verb jumped, 2003: 72). This is responsible for us thinking of a path that starts on the ground, or some other surface, on one side of the hoop. Because of what we know about gravity, we assume that the lady returns to the ground after having passed through the hoop. This endpoint arises in interpretation: it is not coded by the PP itself. (4), too, evokes the idea that an endpoint is reached, namely when je ‘you’ are door Bracciano ‘through Bracciano’. This involves a maximal door-path, i.e. one that goes from one end of Bracciano to the opposite end. Other interpretations, in which the path is shorter, do not seem to be available unless modifiers such as een heel stuk ‘quite a long way’ are used. (4) je kon het [meer] pas zien you could it [the lake] only see



22 A better way to describe this is that de lucht ‘the air’ is unbounded in the scope of predication (cf. Langacker 1987: 203-207 on mass nouns). 23 A difference between my analysis and Tyler and Evans’ (2003) analysis of The cat jumped over the wall is that they analyze English over as a point adposition, while I analyze Dutch over and door as path adpositions.


My conclusion about PreCs with path adpositions is that regardless of which auxiliary is used, hebben or zijn, the prePPs are complements. That is to say, that is what I conclude about the cases of PreCs with path adpositions that are discussed in the analyses discussed in section 5.2. The prePPs in those constructions are complements, because they elaborate a substructure of the motion verb.

I do not want to claim, however, that prePPs with path adpositions are always complements. An example of a modifier prePP with over is (22), from Geeraerts (1992: 207). (22) Waldemar fietst over de lijn Waldemar cycles over the line Geeraerts (1992: 207) distinguishes three readings for (22):

First, Waldemar may be riding across the line (as when he moves from one lane to the other). Second he may be riding on and along the line, staying roughly within its boundaries as he follows it (He might be doing this as a game, or as a kind of test to enable the police to ascertain whether he is riding under the influence of alcohol). [...] And third, Waldemar may be riding at the other side of the line (seen through the eyes of an implicit observer).

It is the third reading that satisfies the constraints imposed by a PreC in which the prePP is a modifier. The process designated by fietst ‘cycles’ elaborates the trajector of the prePP over de lijn ‘over/across the line’. This is a case of subjective motion (see section 4.3.6 and Langacker 2000: Ch. 10); the process takes place at the endpoint of the mental path.24 The first two readings that Geeraerts describes are different interpretations that satisfy the constraints imposed by a PreC in which the prePP is a complement: the subject referent moves along a path that follows a surface of the landmark. It is the context that imposes further constraints: the contexts that Geeraerts describes in brackets narrow down the interpretative possibilities. Returning now to the question whether auxiliary choice can be related to two types of prePPs, we can conclude that that is not possible. Section 5.3 showed that prePPs with path adpositions can occur with hebben

24 See for a similar construction, but without a motion verb, example (i): (i) Nederlanders [...] tanken en kopen [...] steeds vaker over de grens (lr92-4) the-Dutch fill-up [their cars] and buy more-and-more often over the border

‘The Dutch fill up their cars and do their shopping more and more often across the border’

92 Chapter 5

and zijn, while the present section has argued that such prePPs are all complements. This section has also argued that hebben and zijn cross-cut the distinction between Modifier PreCs and Complement PreCs, see especially (6), (7), and (10). Now, if auxiliary choice with manner of motion verbs cannot be related to two types of prePPs, how can these data be accounted for? This question is addressed in the following section. 5.5 Hebben and zijn: different construals My short answer is that it is a matter of construal. Honselaar (1987) accounts for the choice between hebben and zijn in general, i.e. also for other verbs than motion verbs, in terms of the subject referent undergoing some kind of change or not. In the case of manner of motion verbs, hebben is chosen when the focus is on the motion event as a type of act that the subject referent engages in, while zijn is chosen when the focus is on the motion event as the subject referent’s change of location (see also Geerts et al. 1984, Shannon 1990, and Haeseryn et al. 1997). In the case of point adpositions, I distinguish Modifier PreCs and Complement PreCs. With modifier prePPs, see (1) and (10), the motion takes place at a certain location. In (1), the focus is on the type of motion, not on the change of location of the subject referent, and hebben is used: de Veenruiters ‘the Peat Riders’ engage in an act of riding. (1) Zaterdag hebben de Veenruiters Saturday have the Peat-riders in de Pasmanmanege in Steenwijk gereden (MCMAR93SPO) in the Pasman-arena in Steenwijk ridden ‘Saturday the Peat Riders rode in the Pasman Arena in Steenwijk’ (10) Waarom is hij gesprongen in Amsterdam?25 why is he jumped in Amsterdam ‘Why did he jump in Amsterdam?’ In (10), on the other hand, the motion event that takes place in Amsterdam is the following: Dutch musician and painter Herman Brood jumps from the roof of the Amsterdam Hilton and does not survive his fall to the ground. The focus is not on the type of act, but on the subject referent undergoing a change from being on top of a roof to down on the ground.

25 Http://www.condoleance.nl/register.php?id=145&start=1960, March 2008.


Complement prePPs with point adpositions give the endpoint of the motion. In such cases, the change of location construal is a natural one, in which case zijn is used, as in (3). (3) De automobilist A.D. uit Hoensbroek is vrijdag [...] the driver A.D. from Hoensbroek is Friday naast Rijksweg 32 in de sloot gereden (MCAUGSPO) alongside highway 32 in the ditch driven


However, a ‘type of act’ construal is possible as well, which accounts for the use of hebben in (6) and (7). (6) mijn ventje had van de kant in het water gesprongen26 my guy-DIM had from the side in the water jumped ‘my little one had jumped from the side [of the pool] into the water’ (7) Het meisje had in het zwembad gedoken27 the girl had in the swimming-pool dived ‘The girl had dived into the swimming pool’ Remember that (6) is a proud mother’s story of how her little son dared to jump into the pool. He may have done so only once. This change of location is a considerable feat for the boy; the jump into the water is conceptualized, in other words, as a type of act. (7) is a report of a girl who has dived into the pool while it is forbidden to do so. She is asked to leave the pool, and when she refuses, she is eventually arrested by the police. Here too, the subject referent may have dived only once. The change of location is in this case also conceived of as a type of act, be it not an admirable one as in (6), but an illegal one.28 In the case of PreCs with path adpositions, the motion event necessarily involves a subject referent changing location: the prePP

26 Http://tweekids.web-log.nl/twee_kids/index.html, July 2007. 27 Http://zwembadgids.nl/zwembadnieuws.asp, July 2007. 28 The examples with hebben discussed here all involve volitional acts. As (i) shows, however, constructions with hebben can also be used to describe acts that are not volitional. (i) ik had in stront gestapt en had dat helemaal niet door

(Http://www.riffia.nl/showthread.php?p=26217, September 2008) I had in shit stepped and had that completely not through ‘I had stepped in shit and wasn’t aware of it at all’

94 Chapter 5

elaborates the path that is inherent in the semantic structure of the motion verb.29 In all of the PreCs with path adpositions discussed in this chapter, the ‘other’ auxiliary is possible too, but it has a semantic effect. The use of hebben in (2) ‘downplays’ the change of location that the subject referent is involved in, and focuses on the act of walking that she engages in. It is used as an illustration of the subject referent’s mood, cf. met de lippen stijf op elkaar ‘with her lips tightly pressed together’. The use of zijn in (4), conversely, brings out that the motion is construed as a change of location, namely the path that needs to be traversed before you can see the lake. (2) met de lippen stijf op elkaar had zij





(8) and (9) with hebben differ from (2) in that we know quite precisely which paths the subject referents of (8) and (9) have followed with respect to the landmark, while that is not so in (2). Still, hebben is used in (8) and (9) too, which can be accounted for in the same way as (6). The changes of location in (8) and (9) are feats, i.e. admirable types of acts that the subject referents are able to perform. (8) de juffrouw die met haar paard


29 I am only considering PreCs here that contain a complement prePP. 30 Http://www.dbnl.org/tekst/looy001jaap01, October 2004.



Examples (11), (12), and (14), finally, are very subtle in that hebben

is perfectly possible too. The generalization seems to be that zijn brings out that the change of location is more important than the type of act involved. The speaker that uttered (11) is quite possibly surprised at the distances that the hijackers were able to cover. (12) too evokes an image in which the man covers quite a bit of distance (until he loses sight of the prostitute, see (13) in section 5.3). (11) Hoe kan het dat de kapers twee uur how can it that the hijackers two hours door Amerikaans luchtruim zijn gevlogen through American airspace are flown


(12) Een 25-jarige man uit Maasdijk is dinsdagmiddag in zijn blootje a 25-year-old man from Maasdijk is Tuesday-afternoon in his nudity door het Spijkerkwartier in Arnhem gerend, op zoek naar een through the Spijker-quarter in Arnhem run, in search for a prostituée die […] zijn portefeuille […] had gestolen (mcjun95ove) prostitute who […] his wallet […] had stolen ‘Tuesday afternoon a 25-year-old man from Maasdijk ran naked through the Spijkerkwartier in Arnhem, looking for a prostitute who had stolen his wallet’

In (14) too, the focus seems to be not so much on whether the subject referents have engaged in the act of driving over the landmark, but whether they ever followed a path over it.

31 Http://www.kvac.be/clubblad/490204.htm, October 2004. 32 ‘Het complot van 11 September [The 9/11 conspiracy]’, Zembla, VARA/NPS, September 10, 2006.

96 Chapter 5

(14) Maar of [... koningen, prinsen en prinsessen en hun gevolg] ooit But whether kings, princes, and princesses and their retinue ever over de weg van Ruinen naar Echten zijn gereden is niet bekend (mcdec95ove) over the road from Ruinen to Echten are ridden is not known ‘But whether kings, princes, and princesses and their retinue ever rode along the road from Ruinen to Echten is not known.’

In summary, then, auxiliary choice cannot be related to two types of PreCs. Instead, hebben construes the motion event as a type of act, while zijn construes it as a change of location. 5.6 Conclusion This chapter underscores the importance of using attested data. Analyses that relate auxiliary choice to two types of prePPs cannot account for the diversity of actual language use. The chapter also underscores the importance of the notion construal. The choice between hebben and zijn depends on how a motion event is conceptualized: as a type of act (hebben) or as a change of location (zijn). The chapter has paved the way for a comparison between the semantics of PreCs and ParCs, which is the topic of the following chapter. PreCs are of two types: those with modifier prePPs and those with complement prePPs. Chapter 6 proposes how ParCs differ semantically from these two types of PreCs.

Chapter 6 Constructional semantics: ParCs vs. PreCs 6.1 Introduction This chapter constitutes the heart of the present study: it presents an analysis of the semantic differences between ParCs and PreCs, which forms the basis for the analysis of the grammatical constituency of ParCs in Chapter 7. The present chapter analyzes the semantics of the ParC in terms of a single constraint, just as Chapter 4 analyzed the semantics of a word, Dutch over, in terms of a single constraint. The constraint is a generalization over the semantic structures of such diverse ParCs as (1)-(6), which consist of different nominals, verbs, and adpositions. (1) Johan Cruijff sprintte het veld op (jgdmay92)

Johan Cruijff dashed the field on/up ‘Johan Cruijff dashed onto the field’

(2) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’ (3) hun fietspontje [vaart] het kanaal over (jgdapr95)


(4) Ze […] liep de zaal door (de Winter 1989: 343-344) she walked the room through ‘She crossed the room’

(5) [Hij] huppelde […] het Piazza Navona af (de Winter 1989: 305) he skipped the Piazza Navona off ‘He hopped and skipped off the Piazza Navona’

(6) Ze lopen […] het dorp uit (Enquist 1996: 80) they walk […] the village out ‘They are walking out of the village’

The chapter argues that what ParCs share semantically is the

following constraint:

98 Chapter 6

(7) The Dutch ParC profiles a motion event in which a trajector traverses1 a landmark so that result P is achieved: the trajector moves from where it is not P to where it is completely P.2

I assume, along the lines of the cognitive-grammar view of grammatical constructions (see section 5.4.1), that speakers of Dutch acquire specific constructions such as (1)-(6) first, from which they abstract a ParC schema. The constraint in (7) is the semantic pole of this schematic ParC’s composite structure.

The details of the characterization are explained in the chapter, but two remarks are in order here. The first is that ‘P’ in (7) stands for the spatial relation designated by the semantic structure of the adposition. ‘Motion event’, secondly, is used in the sense of Talmy (1985: 60-61):

The basic motion event consists of one object (the ‘Figure’) moving or located with respect to another object (the reference-object or ‘Ground’). It is analyzed as having four components: besides ‘Figure’ and ‘Ground’, there are ‘Path’ and ‘Motion’. The ‘Path’ [...] is the course followed or site occupied by the Figure object with respect to the Ground object. ‘Motion’ [...] refers to the presence per se in the event of motion or location (only these two motion states are structurally distinguished by language).

Given the cognitive-grammar background of this dissertation, however, I use Langacker’s terms ‘trajector’ and ‘landmark’ instead of Talmy’s ‘Figure’ and ‘Ground’ in order to achieve terminological consistency.3

Throughout the chapter ParCs are contrasted with PreCs to show that they impose different constructional constraints. In particular, the chapter argues that PreCs, such as for example (8)-(10), differ from ParCs in that PreCs need not be resultative, nor need the trajector of a PreC traverse the landmark or end up completely P.

1 ‘Traversal’ is defined here in a special way, which is explained in section 6.2. 2 Ideally, the formulation of the constraint would be such that it shows how the notions in italics are related to one another. What I consider to be the constraint’s essence is that P, i.e. the spatial relation designated by the adposition, is the result of the motion event. The other notions in italics seem to be implications of that. For example, if the result of a motion event is P, then the event must be constituted by a transition from not P to completely P (cf. the quote from Vanden Wyngaerd 2001 in section 6.2.3). For an explanation of how traversal is implied by P as the result of the motion event, see section 7.5.3. 3 Talmy (2000: 253) calls Langacker’s ‘trajector’ and ‘landmark’ notions ‘quite comparable’ to his ‘Figure’ and ‘Ground’.

Constructional semantics: ParCs vs. PreCs 99

(8) De man reed op de Prinsenstraat (mcfeb93ove) ‘The man was driving on Prince Street’ (9) Japie [sprong] op de grond (gp94-2) ‘Japie jumped on the ground’ (10) Het legervoertuig […] reed […] op een mijn (jgdoct93) ‘The army vehicle drove on a mine’

The chapter is structured as follows. Section 6.2 explains the notion traversal, section 6.3 the idea that the trajector moves to where it is completely P, and section 6.4 the notion of result. Section 6.5 discusses seemingly problematic ParCs, namely those with the adverb verder ‘further’ and/or what Helmantel (2002) calls ‘length modifiers’. Section 6.6 summarizes the semantic differences between ParCs and PreCs. 6.2 Traversal 6.2.1. A broad definition Traversal, see the semantic constraint in (7), is defined in a rather broad way in this analysis. What is preserved from the normal meaning of traversal is the idea that a trajector moves with respect to a landmark, i.e. it changes position where the landmark is. What is not preserved, however, is the idea that the trajector necessarily moves from one boundary of the landmark to an opposite boundary. Traversal as used here requires only that the trajector’s path covers some portion of the landmark. What is required, in other words, is that the trajector moves from one point to another point with respect to the landmark. Traversal in this broad sense therefore includes not only cases such as represented in Figs. 1a and 1b, but also cases such as represented in Figs. 1c-h. Each of these figures is a schematic representation of the semantics of ParCs with a specific adposition. The circle represents the trajector of the motion event, the rectangular the landmark, and the arrow the trajector’s traversal of the landmark. Figs. 1c-1f are not cases of traversal in the ordinary sense: the trajector does not move from one boundary of the landmark to the opposite boundary. They are, however, cases of traversal in the broader sense: the trajector’s path covers some portion of the landmark. Nor are Figs. 1g and 1h cases of traversal in the ordinary sense: the trajector does not move across or through the landmark, but rather alongside or around it. Still, they are instances of my broad definition of traversal: the trajector’s path covers a portion of the path, cf. Langacker (1991a) on German um ‘around’ and durch ‘through’: “um resembles durch in that the

100 Chapter 6

path traverses the landmark [...] (albeit on the perimeter rather than through the interior)” (1991a: 402).

I refer to the type of traversal illustrated by Figs. 1a and 1b as complete traversal, which is discussed in section 6.2.2. Complete traversal is required by ParCs with over ‘over, across’ and door ‘through’. I refer to the type of traversal illustrated by Figs. 1c-1f as partial traversal, which is discussed in section 6.2.3. Partial traversal is required by ParCs with op ‘on, up’, in ‘in’, binnen ‘inside’, af ‘off, down’, and uit ‘out’. I refer to the type of traversal illustrated by Figs. 1g and 1h as perimeter traversal, which is discussed in section 6.2.4. Perimeter traversal is required by ParCs with langs ‘along’, voorbij ‘past’, om ‘around’, and rond ‘around’.

Fig. 1c. Op Fig. 1d. In/Binnen

Fig. 1e. Af Fig. 1f. Uit

Fig. 1h. Om/Rond

Fig. 1a. Over Fig. 1b. Door

Fig. 1g. Langs/Voorbij


6.2.2 Complete traversal ParCs with door ‘through’ and over require complete traversal: they designate motion events in which a trajector moves from one boundary of the landmark to an opposite boundary of the landmark, see Figs. 1a and 1b. I consider the arrow in Fig. 2, which is Helmantel’s (2002: 110) representation of such constructions, to be too long: the trajector’s traversal of the landmark designated by the construction spans only the area of the landmark.4 There are ParCs with over that allow interpretations such as Fig. 2, but it is not the right semantic generalization for all ParCs with over.

Consider, for instance, the ParCs in (3), (11), and (12). They all impose the constraint that the trajector moves from one boundary of the landmark to the opposite boundary, yet their specific interpretations differ. (3) hun fietspontje [vaart] het kanaal over (jgdapr95)


(11) Ellen fietst het eiland over (Enquist 1996: 162) Ellen cycles the island over

‘Ellen cycles across the island’ (12) Een andere auto rijdt de kruising over (mcdec93ove) an other car drives the intersection over ‘Another car drives across the intersection’ A likely interpretation for (3) is that the ferry moves from one boundary of the canal to the opposite boundary, as in Fig. 1a.5 A likely interpretation for (11) is that Ellen’s path remains within the boundaries of the island, see Fig.

4 Helmantel’s representation views the motion event from above; mine views the motion event from the side. 5 As will become clear below, a ‘likely interpretation’ is an interpretation that arises in a default or prototypical context.

x

over (x)

Fig. 2. Helmantel’s representation of postPPs with over

102 Chapter 6

3 below. A likely interpretation for (12), finally, is something along the lines of Fig. 2: the car approaches the intersection, traverses it, and drives on. These differences arise because the interpretations must satisfy not only the constructional constraint, but also, at the same time, the constraints imposed by the words used in the construction and the contextual constraints (see section 4.3.2).

In the case of (3), for example, vaart ‘sails’ designates a process of motion in which a trajector moves over water. The trajector is elaborated by the profile of hun fietspontje ‘their bicycle ferry’, whose base includes knowledge that ferries transport people across bodies of water and that they go back and forth. The trajector’s path cannot cross the boundaries of the landmark, het kanaal ‘the canal’, because it needs to be supported by water to be able to sail. Instead it moves from one boundary to the opposite boundary, allowing its passengers to get on and off.

The motion event designated by (11) does not go beyond the boundaries of the landmark either: the kind of motion involved, fietst ‘cycles’, is such that it has to take place on land, not on water.6 Since the landmark in (11), het eiland ‘the island’, is a body of land that is surrounded by water, the trajector’s motion is necessarily confined to the boundaries of the landmark. The context of (11), see (13), imposes further constraints.

(13) Ellen gaat een week met de tweeling naar Terschelling. [...] ‘Ellen takes the twins to Terschelling for a week.’ Op de laatste dag [...] is er een felle oostenwind opgestoken. ‘On the final day a fierce eastern wind has gotten up.’ Ellen fietst het eiland over naar de oostpunt, ‘Ellen cycles across the island to its eastern point,’ waar een huis ligt dat Finisterra heet. ‘where there is a house that is called Finisterra.’ Terschelling is a Dutch island with an elongated shape, see Fig. 3. It is narrow on its north-south axis and long on its east-west axis. It is unlikely that Ellen starts to cycle at its ‘technical’ western boundary, i.e. where the land meets the water, because that is where the beach is. It is more likely that

6 Cycling on water is possible in a pedal boat, but a pedal boat is, obviously, unsuitable for cycling across the island.

Fig. 3. Likely interpretation for (11)


she has rented a house that is located somewhere in the western part of the island and that she starts to cycle from there. It is also unlikely that she traverses the island as far as its ‘technical’ eastern tip: there is, to my knowledge, no bike path that goes as far as the boundary between the land and the water. The constructional constraint that the trajector moves from one boundary of the island to the opposite boundary and the contextual constraint constituted by our knowledge of Terschelling’s ‘cyclable’ parts apply in parallel. The optimal interpretation we construct is one in which Ellen cycles from a position sufficiently near one boundary of the island to a position sufficiently near an opposite boundary of the island, that is, ‘sufficient’ for the motion event still to be construable (or categorized) as going from one boundary of the island to the opposite boundary. In (12), conversely, the trajector’s path most likely goes beyond the boundaries of the landmark. (12) Een andere auto rijdt de kruising over (mcdec93ove) an other car drives the intersection over ‘Another car drives across the intersection’ Here too, the construction imposes the constraint that the trajector moves from one boundary of the landmark to the opposite boundary. Because we think of a default context,7 we assume that the car is moving before it comes to the intersection, and that it drives on after it has traversed it. This satisfies the constructional constraint. If we know, however, that the context is such that the car was waiting for a red light at a boundary of the intersection, the interpretation of (12) will be that the car starts driving from there, moves across the intersection, and drives on. Or, if we know that the road is blocked on the other side of the intersection, the interpretation is such that the car stops when it has crossed the intersection. These are all interpretations that satisfy the constructional and contextual constraints.

Note, finally, that traversal may be fairly minimal, i.e. the landmark may have only very minimal extension.8 This is what we find in examples such as (14) from Helmantel (2002: 74) or (15), discussed in 3.3 and 3.4:

7 Cf. Fauconnier (1994: xxi) on the interpretation of sentences in isolation. 8 That is, traversal per se does not require that the landmark has considerable extension. The particular adposition that is used in a ParC imposes further constraints. I argue in section 6.2.3 that ParCs with in or op ‘on’, in which the trajector traverses the landmark so that the result completely in or op is achieved, do require that the landmark has considerable extension.

104 Chapter 6

(14) Peter rent de klapdeuren door Peter runs the swinging-doors through ‘Peter runs through the swinging doors’ (15) De jongens vluchtten [...] de grens over (jgdmay92) the boys fled the border over ‘The boys fled across the border’ As was pointed out in Chapter 3, Helmantel (2002) proposes that such examples constitute a different type of construction because they defy her generalization that the nominal that precedes the adposition denotes a path; instead, according to Helmantel, they are interpreted as a point on a path. The constraint in (7) is a more precise formulation of an idea presented in 3.4. It was stated there that the non-subject nominal does not ‘denote a path’, but rather, that the subject referent moves along a path with respect to the referent of the non-subject nominal. The constraint in (7) does not specify that the landmark is interpreted as a path. Instead, it specifies that the trajector traverses the landmark, where ‘traversal’ is defined in a broad way: the trajector’s path covers a portion of the landmark. This can account for cases with landmarks of bigger or smaller extension alike. The trajectors in (14) and (15) move from one boundary of their respective landmarks to an opposite boundary. The traversal in each case is very short, but it is a case of traversal nonetheless: the trajector moves from one point of the landmark to another point.9 What this section has argued is that the constructional constraint imposed by ParCs with over is the same across contexts. Example (11) nicely illustrates the idea that the constraint “is not a part of an interpretation, but rather specifies a (set of) condition(s) that an optimal interpretation should meet” (Verhagen 1997: 22). The interpretation of (11) involves a somewhat shorter path than the constraint specifies, due to the

9 It might be objected here that moving from one point of the landmark to another point is impossible in the case of a linear landmark such as de grens ‘the border’ (and see also de streep ‘the line’ in example (33) of section 6.3 below) because it is one-dimensional. As I noted in section 4.3.3, however, I do not believe that such objects are conceptualized as one-dimensional in the mathematical sense. For us to perceive such an entity, it has to have at least two dimensions, cf. also Langacker’s note, also quoted in fn. 14 of Chapter 4, in his description of how we can perceive a black point against a white background (1987: 108, fn. 6):

‘Point’ is not to be understood in the mathematical sense, which would imply zero extension [...]. Instead it is taken as whatever minimal area functions as a quantum of cognitive processing at a given level of organization.


contextual constraints that must be satisfied as well. The examples in this section support the idea that the same constructional constraint gives, in Fauconnier’s (1997: 66) words, “rise to different ‘interpretations,’ depending on what existing configuration”, i.e. the context, it applies to. 6.2.3 Partial traversal While ParCs with over ‘over’ and door ‘through’ require complete traversal of the landmark, ParCs with op ‘on’, in ‘in’, binnen ‘inside’, af ‘off’, and uit ‘out’ require only partial traversal, i.e. traversal of a portion of the landmark. ParCs with op, in, and binnen, see Figs. 1c and 1d above, impose the constraint that the trajector changes location from a boundary of the landmark to where the result is that the trajector is completely op, in, or binnen.10 ParCs with af and uit, see Figs. 1e and 1f above, impose the constraint that the trajector moves from where it is op ‘on’ or in ‘in’ the landmark to a boundary of the landmark, that is where it is completely af or uit. So here too, I think that the arrows in Helmantel’s representations for postPPs with binnen and uit, see Fig. 4, should not go beyond the boundaries of the landmark.11

That the motion events designated by ParCs with in, binnen, and op start at a boundary of the landmark follows from the constraint formulated in (7), repeated here.12 (7) The Dutch ParC profiles a motion event in which a trajector

traverses a landmark so that result P is achieved: the trajector moves from where it is not P to where it is completely P.

10 What is meant by ‘completely P’ is explained in more detail in 6.3. As a result of a motion event designated by a ParC with op, for example, the trajector as a whole needs to be op ‘on’ the landmark, i.e. completely supported by it. 11 Fig. 4a is from Helmantel (2002: 110); Fig. 4b is adapted from Helmantel (1998: 377). 12 Where they end is discussed in more detail in section 6.4.

x

binnen (x)

Fig. 4a

x

uit (x)

Fig. 4b

106 Chapter 6

As specified in section 6.2.1, traversal means that the trajector’s path covers a portion of the landmark. Secondly, for result P, i.e. the spatial relation designated by the adposition, to be achieved, the trajector needs to be not in that relation when the event starts, cf. Vanden Wyngaerd (2001: 68) on resultative constructions:

resultatives always involve a transition: the subject [of the resultative predicate] gets into a state s as a result of the activity of the main verb. This presupposes that the subject was not in state s to begin with, and hence that a transition has taken place from ~s to s.

These two aspects taken together, i.e. (i) the trajector’s path covers a portion of the landmark and (ii) the trajector starts at a position where it is not P, necessarily means that in the case of ParCs with op ‘on’, in, or binnen ‘inside’, the motion event starts at a boundary of the landmark. At a boundary, the trajector is not yet op, in, or binnen the landmark, but when it starts moving from the boundary to a position where it is op, in, or binnen the landmark, its path covers a portion of the landmark. Just as we saw for ParCs with over, interpretations of ParCs with op, in, and binnen differ with respect to where the motion starts due to lexical and contextual constraints. What (1) requires, for example, is that Johan Cruijff moves from a boundary of the field, which is a clear line. (1) Johan Cruijff sprintte het veld op (jgdmay92)


The construction is from a report of a football match with Cruijff in the role of head coach of FC Barcelona. Coaches are not allowed to stand on the line of the field; they are not even allowed to stand close to that line. Given all this, a likely interpretation is one in which what is profiled by the construction, i.e. motion onto the field from one of its boundaries, is augmented with background knowledge that Cruijff originally started moving from somewhere outside the field. A potential problem for the idea that the trajector moves from where it is not P is that (1) also allows a type of interpretation in which Johan Cruijff is already on het veld ‘the field’ when he starts to move. This is a type of interpretation that Helmantel (2002: 73) observes with respect to construction (16), i.e. that the man in (16) may start to climb from a position where he is already on the ladder.


(16) De man is de ladder op geklommen the man is the ladder on climbed ‘The man has climbed up the ladder’

It is not the case, however, that the trajector can start anywhere on the landmark in such cases: it has to be close to a boundary.13 For (1) to be used in a felicitous way, Johan Cruijff needs to be close to one of the lines of the field when he starts to run. Similarly in (16), the man needs to be on one of the ladder’s lowest rungs when he starts to climb.14 A possible analysis of such interpretations is therefore the following: the trajector’s starting point is so close to a boundary of the landmark that it is construed as located at that boundary, i.e. still not quite op ‘on’ the landmark. I will come back to similar cases in section 6.5. The generalization that the trajector moves from a boundary of the landmark in the case of ParCs with op, in, and binnen is supported by data with what Helmantel (2002) calls ‘length modifiers’, such as 5/vijf meter ‘five meters’: (17) je hoeft maar 5 meter de straat op te gaan you have just 5 meters the street on/up to go

en je hebt een huwelijksaanzoek te pakken15 and you have a marriage-proposal to grab

‘All you have to do is go five meters out onto the street, and you have got yourself a marriage proposal’

(18) Ik was nog geen vijf meter de straat in gereden I was yet no five meters the street in driven

of twee agenten hielden me al aan16 or two officers held me already on ‘I had not driven five meters into the street when two officers stopped me’

As Helmantel puts it, “[t]he modifier gives the value of the length of [the path] relative to the object” (1998: 378), see her representation of 2 meter x in ‘two meters into x’ in Fig. 5. What the modifiers in (17) and (18) specify, in other words, is the length of the path reckoned from the boundary of the

13 Motion events that do not start close to a boundary can be expressed by means of ParCs with the adverb verder ‘further’, which are discussed in section 6.5. 14 I am assuming that the ladder is in an upright position. 15 Http://marokko.judidt.nl/?p=36, November 2007. 16 Http://www.suuscommunicatie.nl/nieuws/index.php, November 2007.

108 Chapter 6

landmark, which is where the motion events designated by these constructions start.

Motion events designated by ParCs with af ‘off, down’ and uit ‘out’, conversely, start somewhere in or op ‘on’ the landmark and end at a boundary of the landmark. (5) [Hij] huppelde […] het Piazza Navona af (de Winter 1989: 305)

he skipped the Piazza Navona off ‘He hopped and skipped off the Piazza Navona’


This too follows from the constraint in (7). If the result of the motion event is that the trajector is af ‘off’ or uit ‘out’, it needs to be op ‘on’ or in ‘in’ when the event starts (cf. the quote by Vanden Wyngaerd above).17 A

17 Examples such as (i) and (ii) are a possible exception to this: (i) Vergeet je jas niet, als je morgen de deur uit gaat (jgdjun95) forget your coat not if you tomorrow the door out go

‘Don’t forget your coat if you’re going out (lit. out of the door) tomorrow’ (ii) De vrouw werd [...] zo bang dat zij het raam uit sprong

(http://www.nieuwsbank.nl/inp/2005/02/28/R237.htm, June 2008) the woman got so frightened that she the window out jumped

‘The woman got so afraid that she jumped out of the window’ It does not seem likely that the trajector in such cases is first in the landmark (de deur ‘the door’ or het raam ‘the window’). Instead, the non-subject nominals in (i) and (ii) appear to be understood metonymically, i.e. they stand for the room or space that they provide an exit or entry to. Helmantel (2002: 74) excludes an example like (ii) from her analysis that the nominal preceding the adposition denotes a path. She treats it, in other words, as similar to example (14) repeated here as (iii) (from Helmantel 2002: 74). (iii) Peter rent de klapdeuren door Peter runs the swinging doors through ‘Peter runs through the swinging doors’

2m

Fig. 5. Helmantel’s representation of 2 meter x in


boundary of the landmark marks the end of the event: that is where the trajector is af or uit, and moving further would no longer be a case of traversal. Here too, the actual interpretations may involve paths that stop near, rather than on a boundary of a landmark, or that continue even further.18 Traversal need not involve contact between trajector and landmark: a trajector may cover a portion of the landmark without being in contact with it (see also section 6.2.4 on perimeter traversal). What is required is that the trajector is, if not in contact, then somehow close to or involved with the landmark throughout the motion event. A case of complete traversal without contact is (19), a ParC with over. The trajector is involved with the landmark throughout the motion event because it is drawn to the landmark by gravity.

I consider (iii) to be different from (i) and (ii). While (iii) can be accounted for in a straightforward way by the general semantic constraint that I propose, (i) and (ii) can be accounted for in terms of that constraint in combination with the principle of metonymy (see also fn. 6 of Chapter 9). 18 Length modification in ParCs with af and uit does not present a clear picture and therefore deserve further study. There are cases that can unproblematically be understood as modifying the length of the path within the boundaries of the landmark, see e.g. (i) below. Other examples, see e.g. (ii), seem to allow another type of interpretation in addition to the ‘unproblematic type’ just described for (i). The problematic type of interpretation is one in which the length modifier seems to modify a path outside the boundaries of the landmark. A possible explanation for such an interpretation could be that het plein af ‘off the square’ is understood as ‘when they were off the square’. (i) Intussen waait door een windvlaag de deksel van zijn pannetje en In-the-meantime blows by a wind-blow the lid of his pan-DIM and het ding rolt twee meter de berg af. the thing rolls two meters the mountain off (http://www.berggeit.nl/grote-hoogten, November 2007)

‘A gust of wind in the meantime blows the lid off his little pan, and the thing rolls two meters off the mountain’.

(ii) Chip en Victor liepen een stukje het plein af Chip and Victor walked a bit-DIM the square off om te kijken of de auto ook echt doorreed. to see if the car also really through-drove Na een paar minuten kwamen ze terug. after a few minutes came they back

‘Chip and Victor walked a short way off the square to see whether the car was actually driving on. After a few minutes they came back.’

110 Chapter 6

(19) Dit meisje [...] heeft al aangekondigd this girl has already announced dat zij volgend jaar gaat proberen that she next year goes try om ook Amerika over te vliegen (jgdsep93) to also America over to fly

‘This girl has already announced that next year she will try to fly across America as well’

Nor do the ParCs with op ‘on’ in (20) and (21) require that the trajector is in contact with the landmark during the traversal. (20) De fanatieke fan sprong het veld op19 the fanatical fan jumped the field on/up ‘The fanatical fan jumped onto the field’ (21) Ze sprong de bar op20

she jumped the bar op ‘She jumped onto the bar’

The interpretations of (20) and (21) that I am concerned with here are those in which the trajector ends up on the landmark as the result of a single jump. During that jump, the trajector is not in contact with the landmark; only at the end of it is the trajector in contact with the landmark, i.e. completely supported by it.

What is required, however, is that the trajector moves close to the landmark, traversing some portion of the landmark. Figs. 6a and 6b represent these interpretations of (20) and (21). (20) profiles a motion event, see Fig. 6a, in which the trajector traverses a portion of the landmark by means of a single jump; only in the final state of the jump, when the fan lands on the field, is contact established. (21) similarly profiles a motion event, see Fig. 6b, in which the trajector traverses a portion of the landmark, for example from somewhere close to the bottom of the landmark the bar to the top, by means of a single jump. In each case, the trajector’s path covers a portion of a surface of the landmark.

19 Http://www.voetbalzone.nl/doc.asp?uid=44197, November 2007. 20 Tropical Danny & DJ Guillermo, Toppertje, 2006.


If we now contrast ParCs with PreCs, we can observe that the

requirement of (partial) traversal in ParCs is absent in PreCs with op (or in). The trajector’s path in the case of such PreCs may not cover some portion of the landmark. This is why the PreC in (22) is acceptable, but its ParC counterpart in (23), from Schermer-Vermeer (1970: 97) is not. (22) Japie [sprong] op de grond (gp94-2) ‘Japie jumped on the ground’ (23) *Hij sprong de grond op he jumped the ground on/up All that the PreC in (22) requires is that Japie jumps from a position where he is not op de grond ‘on the ground’ to a position where he is op de grond ‘on the ground’. This is satisfied by the interpretation represented schematically in Fig. 7. Japie, represented by the circle, is on the ground, represented by the horizontal line, as a result of a jump from somewhere above the ground (such as from a tree).

Schermer-Vermeer (1970: 97) explains the ungrammaticality of (23) by saying that the motion designated by a ParC with op needs to be upward, which is in fact a common assumption (cf. Kraak and Klooster 1968: 223, Luif 2000: 191). The data presented here show that this cannot be the right generalization: the motion profiled by a ParC with op may be horizontal, as

Fig. 7

Fig. 6a

Fig. 6b

112 Chapter 6

in (1), or even slightly downwards as in (20) or (24), for example when the surface of the ice (rink) is somewhat lower than its surroundings. (1) Johan Cruijff sprintte het veld op (jgdmay92)


(20) De fanatieke fan sprong het veld op21 the fanatical fan jumped the field on/up ‘The fanatical fan jumped onto the field’ (24) Ik ben vol vertrouwen het ijs opgestapt (mcdec95spo)

I am full-of confidence the ice on-stepped ‘I stepped onto the ice full of confidence’

The problem in (23) is the landmark de grond ‘the ground’: it does not have a clear boundary from which the trajector can start its (partial) traversal of the landmark from not P to completely P. 22 Het ijs ‘the ice’ in (24) is different in this respect: the speaker is a skater, which creates a context in which het ijs ‘the ice’ is construed as having a clear boundary, such as for example in the case of an ice rink. Compare also the difference between (25) with de vloer ‘the floor’ and unacceptable (26) with de grond ‘the ground’. 21 Http://www.voetbalzone.nl/doc.asp?uid=44197, November 2007. 22 De grond ‘the ground’ is similar to bare mass nouns in this respect, which Helmantel (2002: 74-75) does not expect to find in postPPs, cf. the attested PreC in (i) and the semantically anomalous ParC in (ii). (i) er was een vrijwilligster die vastbesloten was there was a volunteer who determined was

op glas te springen en dat ook deed (mcmar94ove) on glass to jump and that also did

‘There was a volunteer who was determined to jump on(to) glass and who in fact did that’

(ii) * ze was vastbesloten glas op te springen she was determined glass on/up to jump Example (iii), however, shows that bare mass nouns can occur as the non-subject nominal of a ParC. (iii) Wat is er nou voor leuks aan om massaal ijskoud water in te duiken? (http:// www.radionof.nl/?page=1&id=4704&view=reactie, May 2008)

what is there now for fun on to all-together ice-cold water in to dive ‘What is the fun in diving all together into freezing water?’ The explanation for the occurrence of the bare mass noun in (iii) is that the water has a surface. This boundary between the water and the air above it is the boundary from which the traversal can take place.


(25) Simon Vinkenoog liep de vloer op23 Simon Vinkenoog walked the floor on/up ‘Simon Vinkenoog walked onto the floor’ (26) *Simon Vinkenoog liep de grond op Simon Vinkenoog walked the ground on/up An important difference between ParCs and PreCs with op, therefore, is that ParCs require (partial) traversal, while PreCs with op do not. 6.2.4 Perimeter traversal The types of traversal that have been discussed so far consisted in the trajector’s change of location ‘covering’ some portion of the landmark: by being in contact with the landmark, or by being drawn to the landmark by gravity, or somehow close to the landmark. The latter type is also found in the ParCs discussed in the present section, with langs ‘along’, voorbij ‘past’, om ‘around’, and rond ‘around’, see (27)-(30). (27) De kinderen gaan de huizen langs24 the children go the houses along ‘The children are going by the houses’ (28) Ik [...] liep de bloemenzaak voorbij25 I walked the flower-shop past ‘I walked past the florist’s shop’ (29) hij liep de hoek om (Enquist 1996: 312) he walked the corner around ‘he walked around the corner’ (30) Bob en Sally gaan de wereld rond (Enquist 1996: 187) Bob and Sally go the world around Bob and Sally are going (on a trip) around the world Such ParCs require what I call ‘perimeter traversal’, cf. also Langacker (1991a) on German um ‘around’ and durch ‘through’: “um resembles durch in that the path traverses the landmark [...] (albeit on the perimeter rather than through the interior)” (1991a: 402).

My discussion of ParCs with langs ‘along’, voorbij ‘past’, om ‘around’, and rond ‘around’ is only very limited; they require further study. 23Http://www.cubra.nl/specialebijdragen/BrabantLiterair/brabantliterairpdfbestanden/BCBL_Jan_2008_Jasper%20Mikkers_Hans%20Vlek_essay.pdf, May 2008. 24 Http://www.travelpoort.nl/s_6_37_14.html, April 2008. 25 Http://prlwytskovsky.punt.nl/?a=2007-03, April 2008.

114 Chapter 6

My hypothesis, however, is that the constraint formulated in (7) is the right generalization for these ParCs as well. Consider the representations in Figs. 1g and 1h, which present views from above.

Fig. 1g, for ParCs with langs and voorbij, represents a motion event in which a trajector moves along a side of the landmark from one end to the other. Fig. 1h, for ParCs with rond and om, represents a motion event in which the trajector moves all around the landmark. What the paths in Figs. 1g and 1h share with those of ParCs with other adpositions is that the trajector moves while, if not in contact with, then close to the landmark, i.e. the trajector moves only there where the landmark is.

Further study should reveal how, for one thing, ParCs with langs and voorbij are different semantically. One representation for two types of ParCs does not do justice to, for example, the intuition that a ParC with langs such as (27) evokes the idea of some sort of interaction with the houses: the children are going from door to door, for instance because they are collecting for some charity. This idea of interaction is not at all evoked in a ParC with voorbij such as (28): the objective seems to be to get past the houses. Similarly, just one schematic representation for ParCs with rond and om does not seem correct either. ParCs with rond, for example, also allow interpretations, not captured by Fig. 1h, in which the shift of location takes place within the boundaries of the landmark (see Zwarts 2006 for a recent study of Dutch om and rond, and Dewell 2007 on English around). 6.3 Completely P The trajector’s traversal profiled by a ParC results in the relation completely P. What this means depends, obviously, on the adposition involved. In the case of a ParC with op ‘on’, for example, the result of the motion event is that the trajector is completely op (the landmark), which means that the trajector is completely supported by the landmark and that no part of the trajector is supported by something else. In (1), for example, the result of the motion event is that Johan Cruijff is completely op, i.e. supported by, the

Fig. 1h. Om/Rond

Fig. 1g. Langs/Voorbij


landmark. In (2), similarly, the trajector T. is, as the result of the motion event, completely in, i.e. contained in, the street: no part of T. is outside, i.e. not contained by, the street. (1) Johan Cruijff sprintte het veld op (jgdmay92)


(2) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’ It is this semantic difference between ParCs and PreCs combined with the different requirements with respect to traversal that explains the difference in acceptability between (31) and (32), from Helmantel (2002: 77). (31) Jan stapt op de kiezelsteen ‘Jan steps on the pebble’ (32) #Jan stapt de kiezelsteen op Jan steps the pebble on (32) does not allow interpretations in which Jan is a human being of normal height, and the pebble has the size pebbles normally have. It only allows interpretations in which Jan is very small, or the pebble very large (Helmantel 2002: 77, cf. also Schermer-Vermeer 1970: 97). Helmantel accounts for the difference in acceptability between (31) and (32) in terms of her generalization that the nominal that precedes the adposition denotes a path, but section 3.4 argued that this generalization is problematic. Instead, it is the ParC’s requirement that the trajector traverses the landmark from not P to completely P that accounts for the semantic anomaly of (32). As pointed out in section 6.2.2, the notion traversal per se does not require that the landmark has considerable extension, cf. the discussion of (14) and (15) above. Example (33), which is a ParC with over, illustrates this point: the landmark de witte streep ‘the white line’ has hardly any extension, yet it is sufficient for the trajector to traverse it. (33) [...] stapte ik de witte streep over26 stepped I the white line across I stepped across the white line

26 Http://books.google.nl, June 2008.

116 Chapter 6

Why then is (32), in which the landmark is also of limited extension, not acceptable? The semantic anomaly of (32) is due to the requirement that the result of the traversal is that the trajector is completely P. The trajector Jan in (32) needs to move from a boundary of the landmark de kiezelsteen ‘the pebble’ to where he is completely op ‘on’ the landmark, i.e. completely supported by the pebble and nothing else. What satisfies this constraint is an interpretation that features, for example, a person Jan who is as small as an ant or a pebble the size of a boulder (Helmantel 2002: 77). Because these are non-default contexts, the construction is strange. The PreC in (31) is fine: it does not require that the trajector ends up completely supported by the landmark. Probably only a small part of Jan’s foot is in contact with the pebble: the rest of Jan is supported by something else, e.g. the ground. The ParC with over in (33), finally, is fine too: it does not require that the trajector ends up completely supported by the landmark de witte streep ‘the white line’; it would not have the extension for that either.27 Instead, it requires that the trajector ends up completely over the line, i.e. on the other side of it. Another pair that can be accounted for in terms of this difference between ParCs and PreCs is this: (34) Op 3 december [...] voer een schip [...] op de rotsen bij La Coruña (jgddec92) on 3 December sailed a ship on the rocks near La Coruña ‘On December 3 a ship ran (lit. sailed) on the rocks near La Coruña’ (35) ??Het schip voer de rotsen op the ship sailed the rocks up ‘The ship sailed up the rocks’ The PreC in (34) is fine. It designates a motion event in which the ship sails and then ends up in contact with the rocks. The motion does not constitute traversal of the rocks: only at the end of the motion is the ship in contact with the rocks. The motion precedes, in other words, the contact with the rocks. Nor does the PreC require that the ship is completely supported by the rocks: the support is probably only partial.

27 Cf. the following pair: a PreC and a ParC with op. (i) Wilson stapte op de lijn

(http://zzleidenbasketball.nl/blog/show/264, June 2008) ‘Wilson stepped on the line’ (ii) ??Wilson stapte de lijn op Wilson stepped the line up


The ParC in (35), on the other hand, requires that the ship traverses the rocks so that the result is that it is completely op ‘on’ the rocks, i.e. completely supported by them. This means that the ship must start at a boundary of the rocks and then sail on, traversing the rocks, so that it ends up completely supported by the rocks. This is in conflict with our knowledge of what happens when a ship hits the rocks: it does not sail on over the rocks, but comes to a stand-still.

However, as soon as we think of a context in which such traversal is possible, the ParC loses its ‘strangeness’. (36) is such an example, from a context in which the trajector (a tow boat) is not a real physical object, but a virtual one, part of a computer program on a naval simulator. The simulator has a flaw: due to a programming error the tow boat does not come to a stand-still when it hits the landmark, but moves on. (36) de sleepboot [kruipt ...] moeiteloos de kant op (jnlmay94) the tow-boat creeps effortlessly the side up ‘The tow boat is effortlessly creeping up the side’ As a final example of the completely P vs. partially P, consider ParC (3) and PreC (37) with over. (3) hun fietspontje [vaart] het kanaal over (jgdapr95)


(37) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrockcanal ‘We are sailing along the Haren-Rüttenbrock canal’ The ParC in (3) requires that the trajector, their bicycle ferry, is completely over as a result of the motion event. We have to construe, in other words, the semantic constraint for over, see Chapter 4, in such a way that its trajector is completely over as a result of a motion event. This means that the trajector has completed a maximal over-path with respect to the landmark, i.e. from one boundary to an opposite boundary. Such a requirement is absent in the PreC in (37): it requires only that the trajector moves along a path that follows a surface of the landmark, as was argued in Chapter 4.

118 Chapter 6

6.4 Result While earlier studies characterize the adpositions in ParCs as directional (e.g. Haeseryn et al. 1997, Helmantel 2002), I propose, see the constraint in (7), that they are understood as the result of the motion event. (7) The Dutch ParC profiles a motion event in which a trajector

traverses a landmark so that result P is achieved: the trajector moves from where it is not P to where it is completely P.

‘Result’ should not be taken to mean ‘intended result’, i.e. the trajector need not intentionally move in a certain way so as to achieve the result completely P. Instead, it is the language user who conceptualizes the trajector’s motion as leading to this result. The trajector need therefore not be animate, see (38) and (39).28 And if the trajector is animate, (s)he need not intend for the motion event, including its result, to happen, see (40), uttered by a cyclist competing in the Tour de France. (38) het water [liep] de huizen in (jgdsep94) the water ran the houses in ‘the water flowed into the houses’ (39) Tulpen komen de grond uit29 tulips come the ground out ‘Tulips are coming out of the ground’ (40) Ik [...] dook [...] 30 meter een ravijn in30 I dove 30 meters a ravine in ‘I dove 30 meters into a ravine’

ParCs are resultative, while PreCs are not necessarily so, cf. the

contrast between (1) and (8): (1) Johan Cruijff sprintte het veld op (jgdmay92)


28 Notice that while de grond ‘the ground’ cannot be used in the ParCs with op ‘on’ in (23) and (26), its use in the ParC in (39) with uit ‘out’ is fine. Just as ijskoud water ‘ice-cold water’ in example (iii) of fn. 22, de grond ‘the ground’ has a surface that functions as the necessary boundary between being not uit ‘out’ to being completely uit ‘out’. 29 Http://www.destentor.nl/regio/flevoland/article1020178.ece, June 2008. 30 NOS teletekst, July 19, 2002.


(8) De man reed op de Prinsenstraat (mcfeb93ove) ‘The man was driving on Prince Street’ The trajector in (1), a ParC, moves from not P, i.e. not op het veld ‘not on the field’, to where it is op het veld ‘on the field’. The trajector in (8), a Modifier PreC (see Chapter 5), on the other hand, is op de Prinsenstraat ‘on Prince Street’ throughout the profiled process.

The notion result is crucial to an account of how far the trajector of a ParC with op or in, traverses the landmark, which is a question that Helmantel (2002) leaves unanswered (see section 3.4). In ParCs with op and in, the traversal may only be very short, see (24) and (41): (24) Ik ben vol vertrouwen het ijs opgestapt (mcdec95spo)

I am full confidence the ice on/up-stepped ‘I stepped onto the ice full of confidence’

(41) dat ie nu de lift in kan stappen that he now the elevator in can step zonder dat hem het angstzweet uitbreekt (mcmar94ove) without that him the fear-sweat out-break

‘that he can now step into an elevator without breaking out in a cold sweat’

In these examples, only one step suffices for the trajector to move from a boundary of the landmark to where it is completely supported by or contained in the landmark.

In other ParCs with op and in, the path may be longer. But how long can it be? Consider example (2) and Fig. 8. (2) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’

Fig. 8a Fig. 8b

Fig. 8d Fig. 8c

Fig. 8. Paths relative to a street

120 Chapter 6

Section 3.4 raised the question how we can account for the intuitions that Figs. 8a and 8b are readily acceptable as representations of interpretations for (2), while Figs. 8c and 8d are not. It was also observed there that if de Leliestraat ‘Lily Street’ in (2) is a dead-end street, interpretations such as represented in Figs. 8c and 8d suddenly become possible. These intuitions and contextual effects can be accounted for by the constraint in (7). In a ParC with in, the trajector can traverse the landmark as long as it is construed as heading towards the result completely in. Without any special context, i.e. without further knowledge about de Leliestraat ‘Lily Street’, we assume that the street can be entered on one end and exited on the other. With respect to such a landmark, the trajector can move in the ways of Figs. 8a and 8b and still be considered to be moving so that the result is that it is in. In Figs. 8c and 8d, however, the trajector’s motion is at some point no longer construable as leading to the result in, but rather as leading to the result not in, i.e. uit ‘out’. This is why Figs. 8c and 8d are not acceptable as interpretations for (2) when assuming a default context, cf. the answer to the riddle in (42) mentioned in section 3.4. (42) Q: Hoe ver kan een hond het bos in lopen? how far can a dog the wood in walk ‘How far can a dog walk into the wood?’ A: tot de helft, daarna loopt de hond het bos weer uit31 to the half then walks the dog the wood again out ‘To the middle, after that the dog is walking out of the

wood’

As soon as, however, there is no expectation that the trajector will leave the landmark, interpretations such as represented in Figs. 8c and 8d are possible for (2). If we know that the street is a dead-end street, there is no expectation that the trajector will leave it on the other side. In that case, the trajector’s path can go all the way to the other end of the street and still be construed as leading to the result in. Similarly, if het bos ‘the wood’ in (42) is surrounded by a high brick wall with only one opening through which the dog enters, it can walk all the way to the opposite end of the wood: there is no expectation that he will leave the wood.

The landmark need not be physically closed off for such longer paths to be allowed, see (43), from a personal conversation.

31 Http://www.kantjils.nl/KK/KK_2002_jun.pdf, January 2007.


(43) We zijn nu veel te ver het dorp ingereden we are now much too far the town in-driven ‘We have driven into the town much too far’ The context of (43) is the following. A group of friends are trying to find the new house of another friend in a town that is unfamiliar to them. They enter the town on one end and follow, roughly, the path that is represented by the arrow in Fig. 9, without success: they do not find the house. When they are nearing the town’s opposite boundary, i.e. opposite from the one where they entered the town, see the arrow’s head in Fig. 9, one of the friends utters the ParC in (43). The complete path that has come before is, in other words, construed by him as a case of het dorp inrijden ‘driving into the town’, be it veel te ver ‘much too far’. This is possible because they have no intention of leaving the town: they are looking for someone’s house which is located in that town. At no point of the path is the motion construed as heading for the result not in.

Similar observations can be made for ParCs with op ‘on’. If the landmark and further context are such that the trajector could be expected to leave the landmark, the trajector’s motion cannot go further than, roughly, the center area of the landmark. In the case of (44) below, however, the landmark is a small ferry, and the trajector Pluk is the driver of a small tow truck. The sizes of the truck and the ferry are such that when the truck is completely op ‘on’ the ferry, its nose is close to the other end of the ferry, i.e. the boundary that is opposite to the one that it started from. It moves, in other words, further than the center area of the landmark. (44) Pluk […] reed voorzichtig het pontje op (Schmidt 1979: 102)

Pluk drove carefully the ferry-DIM op ‘Pluk drove carefully onto the small ferry’

Fig. 9.

122 Chapter 6

Consider also (45) and Fig. 10. Section 3.4 raised the question how we can account for the observation that Figs. 10a and 10b are possible interpretations for (45), but Figs. 10c and 10d are not. (45) De tram reed een heuvel op (jnlmar92)

the tram drove a hill up/on ‘The tram drove up a hill’

The constraint in (7) can account for this observation. Figs. 10a and 10b represent paths that can easily be construed as leading towards the result that the tram is completely supported by the hill. What ‘goes wrong’ in 10c and 10d is that at some point, the motion is no longer oriented towards the result op, but to the result not op ‘on’, i.e. when it starts moving downwards. For that type of motion, geared towards the result not op ‘on’ a different ParC is used, namely with af ‘off’. The examples discussed in this section are, in my view, particularly nice illustrations of Fauconnier’s (1997: 66) idea that

the same ‘meaning,’ conceived of as building instructions, can give rise to different ‘interpretations,’ depending on what existing configuration such instructions are applied to, and how they are applied; the ‘how’ comes into play because the building instructions typically underspecify the [cognitive discourse] construction: There can be more than one way to elaborate an existing configuration in conformity with the instructions carried by the linguistic form of the sentence.

Example (2) is an especially clear example of this. It imposes the same constraint, i.e. what Fauconnier calls ‘the same “meaning[...]”’, whenever it is used. The context, i.e. the ‘existing configuration’, also imposes

Fig. 10. Paths relative to a hill

Fig. 10c Fig. 10d

Fig. 10b Fig. 10a


constraints, which is why the same expression, in this case (2), can lead to different interpretations, see the discussion of Fig. 8.

In addition, the examples discussed in this section underscore the importance of the notion construal. The ParC imposes the general constraint that the trajector’s traversal of the landmark results in the relation ‘completely P’. How far the trajector can maximally traverse the landmark depends on whether the motion can still be construed as leading towards ‘completely P’. That depends, as we have seen, on all kinds of factors, including the language user’s expectations. 6.5 ParCs with verder ‘further’ and ‘length modifiers’ This section discusses ParCs with verder ‘further’ and/or ‘length modifiers’, which may at first sight appear to be counterexamples to the general semantic constraint that I propose for ParCs. I suggest, however, that the general constraint is imposed in these cases too, but that the conceptualization of a motion event from not P to completely P is evoked only as part of the base.

Let us first consider two examples with verder (cf. Verkuyl and Zwarts 1992: 385): (46) De boot voer verder de zee op32 the boat sailed further the sea up ‘The boat sailed further onto the sea’ (47) Dus liepen ze nog verder het bos in33 so walked they still further the wood in ‘So they walked still further into the wood’ These examples may appear to be counterexamples to the general constraint, which states the trajector moves from not P to completely P. What I have claimed is that ParCs with in and op profile a motion event that starts at a boundary of the landmark and ends where the result is that the trajector is completely contained in or supported by the landmark. In (46) and (47), however, the trajector does not start at a boundary of the landmark.34 The

32 Http://annemariemarjolein.waarbenjij.nu, May 2008. 33 Http://members.chello.nl/d.vanegdom/page10.html, December 2005. 34 Their contexts make this quite clear: (i) Ook hebben we dolfijnen gespot, de boot had al na 15 minuten een groep

gevonden die enthousiast met de boot mee zwommen! De boot voer verder de zee op om ‘the hole in de rock’ te bekijken.

124 Chapter 6

trajector de boot ‘the boat’ in (46) is already completely supported by the landmark de zee ‘the sea’ when it starts to move. Ze ‘they’ in (47), similarly, are already completely contained by the landmark het bos ‘the wood’ when they start to move. What is also conveyed, however, is that the trajector’s motion is a continuation of an earlier motion event which has started at a boundary of the landmark.35 This is represented in Fig.10, in which the circle represents the landmark. What (46) and (47) profile is the motion event represented by the arrow in bold, which starts at a point where the trajector is in or op ‘on’ the landmark. What the base also includes, however, is the information that this motion event is the continuation of an earlier motion event, which is represented by the (non-bold) arrow that starts at a boundary of the circle

‘We also spotted some dolphins; it took the boat only 15 minutes to find a

group that enthusiastically swam along with the boat! The boat sailed further onto the sea to visit ‘the hole in the rock’.

(ii) Het was een lange tocht naar Bremen. Toen de avond viel, waren ze er nog niet. Ze waren in een groot bos en besloten daar te overnachten. [... D]e haan [...] zag ineens in de verte een lichtje branden. 'Hé!', riep [hij ...], ‘[...] Laten we daar nog naar toe lopen.[’ ...] De ezel, de hond en de kat vonden het een heel goed idee. Dus liepen ze nog verder het bos in.

‘It was a long way to Bremen. When night fell, they were not there yet. They were in a big wood and decided to spend the night there. The rooster suddenly saw a light burning in the distance. “Hey!”, he shouted, “Let’s go there.” The donkey, the dog, and the cat thought that that was a very good idea. So they walked still further into the wood.’

35 As will be illustrated below, it is not necessarily the trajector itself that has engaged in the motion from the boundary of the event. It may be subjective motion (Langacker 2000: Ch. 10), i.e. the language user’s mental scanning of the path that goes from a boundary of the landmark to where one would be in or on the landmark.

Fig. 10


(the landmark). A ParC with verder, in short, evokes as its base a conceptualization in which a trajector has already engaged in a motion event that satisfies the constraint and then continues to do so. It is the continuation that is profiled by a ParC with verder. Just like (46) and (47) with verder and the point adpositions in ‘in’ and op ‘on’, (48) and (49) with verder with path adpositions may also appear to pose a problem for the general constraint: they designate motion events in which the trajector does not start at a boundary of the landmark. (48) Na de mozaïeken rijden we

visit after the mosaics drive we langs de zuidelijke kustlijn verder het eiland over36 along the southern coastline further the island over ‘After the mosaics we drive further across the island along the southern coastline’

(49) Loop verder de Uiterwaardenstraat door37 walk further the Uiterwaarden-street through ‘Walk further through the Uiterwaarden Street’ Here too, however, each construction evokes as its base the conceptualization of a motion event that satisfies the general ParC constraint, which is represented by the dotted arrow in Fig. 11 below: a trajector traverses the landmark so that the result is that the trajector is completely over or door ‘through’, i.e. it moves from one end of the landmark to the opposite end. Further included in the base is the information that the trajector has been engaging in this motion event for a while: it has completed only part of the path because the trip has been interrupted.38 This is represented in

36 Http://www.gorcumsreisburo.nl/Reisverslagen/verslag_cyprus, April 2007. 37 Http://www.zuidelijkewandelweg.nl/architectuur/anderegezicht, May 2008. 38 The wider context of (48) makes it clear that the trip has started at the western tip of the island and is interrupted for a visit to a villa with mosaics. (49) is an example taken from a description of a city walk, see the wider context in (i). The description now and then invites the visitor to pay attention to certain points of interests. The ParCs with verder invite the visitor to continue the motion event that (s)he was engaging in before the interruption. (i) U komt dan weer bij de Uiterwaardenstraat. Ga hier rechtsaf. [Na

o]ngeveer 100 meter ziet u aan de rechterzijde dit contrast in architectuur. [...] Loop verder de Uiterwaardenstraat door. Ter hoogte van de Bernissestraat is wederom een duidelijk contrast in architectuur te zien. [...] Loop verder de Uiterwaardenstraat door en steek de Maasstraat over.

‘Now you are back in the Uiterwaardenstraat. Turn right. After about 100 meters you can see this contrast in architecture on your right. Walk further

126 Chapter 6

Fig. 12 by the solid (non-bold) arrow. And finally, what is also included in the base is that the trajector resumes engaging in the motion event. The ParC with verder profiles a portion of this sequence of events, namely the final portion, where the trajector continues to engage in the motion event. This is represented by the bold arrow in Fig 12. Since the constructions in (48) and (49) do not specify that the motion event was completed, i.e. all the way to the opposite end of the landmark, the bold arrow does not extend that far. What the constructions designate is that the trajector continues to engage in the motion event, not that it ‘executes’ it completely.

Some data with what Helmantel (2002) calls ‘length modifiers’ may appear to be problematic too, in that the trajector does not appear to be moving from where it is not P. I will discuss an unproblematic case first, i.e. the ParC in bold in (50), of which (40) above is a fragment. (50) [I]n de afdaling van de Aubisque miste een renner voor me in the descent of the Aubisque missed a runner before me de bocht. Ik kon hem nog wel ontwijken, maar the turn I could him still avoid, but

maar dook zo 30 meter een ravijn in (NOS Teletekst 19/07/02) I dove like-that 30 meter a ravine in

‘When we were going down the Aubisque, a runner before me missed a turn. I managed to avoid him, but dove 30 meters into a ravine’

through the Uiterwaardenstraat. At the Bernissestraat there is again a clear contrast in architecture. Walk further through the Uiterwaardenstraat and cross the Maasstraat.’

Fig. 12

Fig. 11


The context of the ParC clearly shows that the speaker, a cyclist, is not in the landmark een ravijn ‘a ravine’ when he takes the fall designated by the ParC. The length modifier 30 meter specifies the length of the trajectory calculated from the boundary of the ravine, i.e. the edge where the cyclist goes over. The ParCs in (51)-(53) are different in that the contexts make it clear that the trajectors are already in ‘in’ or op ‘on’ the landmarks when the motion events start. (51) Zonder aarzelen stapt de schildpad op het witte zand ‘Without hesitation the turtle steps onto the white sand’

en waggelt dertig meter het strand op 39 and waddles thirty meters the beach on ‘and waddles thirty meters onto the beach’

(52) Ze zit onder het balkon van één-hoog, ‘She is sitting under the balcony of the first floor,’ op het koude metalen deksel van de vuilnisemmer [...] ‘on the cold metal lid of the garbage bin’ Ze staat op [...] en doet een paar passen langs de muur. ‘She gets up and takes a couple of steps alongside the wall’ Ze loopt een stukje de tuin in40

she walks a bit-DIM the garden in ‘She walks a short way into the garden’ (53) Jezus [werd] bijna het meer [...] ingeduwd.

‘Jesus was nearly pushed into the lake.’ Daarom stapte hij in een bootje ‘That is why he got into a little boat’ en voer een stukje het meer op41

and he sailed a bit-DIM the lake on ‘and sailed a short way onto the lake’ They are problematic because the general constraint that I have formulated for ParCs states that the trajector moves from where it is not P.

What is also clear, however, is that in each case the trajector is still very close to a boundary of the landmark. (51), for example, describes how a turtle emerges from the water, gets into contact with the sand, and moves thirty meters from that position, away from where it was not op het strand ‘on the beach’. The trajector in (52), similarly, moves from a position close

39 Http://gpdhome.typepad.com/ned_ant_achtergronden/2007/07/index, May 2008. 40 Http://www.janvanmersbergen.nl/publicaties_vn, May 2008. 41 Http://www.jufelisabeth.nl/Degelijkenisvandezaaier, May 2008.

128 Chapter 6

to a wall, i.e. a boundary of the garden. In (53), too, the trajector is only just on the lake, near one of its banks. A possible analysis of such examples is therefore that in such cases the trajector’s starting point is so close to a boundary of the landmark that it is construed as located at that boundary, i.e. still not quite in or op the landmark. An example that under one of its readings defies such an analysis, however, is (54): (54) De beer loopt snel een stukje de berg op42 the bear walks quickly a bit-DIM the mountain on ‘The bear quickly walks a short way up the mountain’ It allows interpretations, unproblematic for the proposed analysis, in which the trajector de beer ‘the bear’ is at the foot of the mountain, i.e. at a boundary and moves a little up the mountain. It is also quite possible, however, that the bear is already well onto the mountain, e.g. halfway, and it moves a little up the mountain from there. (54) is taken from a description of how two hikers spot a bear. If we assume that they are already halfway up the mountain and see the bear there, the latter type of interpretation is the only possible one.

I will therefore suggest an alternative analysis, for which I first turn to ParCs with path adpositions and length modifiers. Helmantel (2002: 112) claims that length modifiers are not compatible with such constructions, because the constructions denote a process along a path that has a fixed length: the path ‘occupies’ the entire landmark. Attested examples (55) and (56) show, however, that this claim is incorrect. (55) hij [...] liep ontspannen een stukje de straat door (Peper 2003: 374) he [...] walked leisurely a bit-DIM the street through ‘he leisurely walked a short way through the street’ (56) We [...] lopen [...] een stukje de markt over43 we walk a bit-DIM the market over ‘We are walking a short way across the market’ What I propose is that just like ParCs with path adpositions and verder, see (48) and (49) above, examples (55) and (56) evoke the conceptualization of a trajector moving from one end of the landmark to the opposite end, as 42 Http://www.sanmarko.nl/nl/Home-/Vakanties_/Alaska/Homer_-_Anchorage.aspx, May 2008. 43 Http://www.landenweb.net/zuid-afrika/reisverhalen/reisverslagzuidafrika, May 2008.


represented in Fig. 11. Een stukje ‘a short way, a little, a bit’ modifies this process: it conveys that only part of the process is ‘executed’, just as in constructions such as (57) and (58), which are not ParCs, but SCVCs (constructions with a separable complex verb that takes a direct object).44,45 (57) een ballon een klein stukje opblazen46 a balloon a little bit-DIM up-blow ‘blow a balloon up a little’

44 Een stukje ‘a bit’ does not, in other words, modify the non-subject nominal; it modifies the process. (55) and (56) are, in other words, different from: (i) Hij liep ontspannen een stukje van de straat door. he walked leisurely a bit-DIM of the street through ‘He leisurely walked through a portion of the street’ (ii) We lopen een stukje van de markt over. we walk a bit-DIM of the market over ‘We are walking across a bit of the market’ In such cases een stukje van X ‘a bit of X’ constitutes the landmark which is completely traversed: the trajector moves from one end of a portion of X to the opposite end of that portion.

While the difference between (55)-(56) vs. (i)-(ii) is very subtle, it is clearer in the case of (57) en (58) vs. (iii) and (iv): (iii) ??een klein stukje van de ballon opblazen a little bit-DIM of the balloon up-blow ??‘blow up a little bit of the balloon’ (iv) ??een stukje van je longen opschonen a bit-DIM of your lungs up-clean ??’cleanse a little piece of your lungs’ In (iii) en (iv) only part of X, i.e. the balloon and the lungs, is involved in the event, while in (57) and (58) it is the events that are carried out only partially: the whole balloon and the whole lungs are partially blown up/cleansed.

In short, in the case of (i)-(iv), een stukje van X ‘a bit of X’ constitutes the landmark of the process, while in (55)-(58) een stukje ‘a bit’ designates that only part of the process is carried out. 45 In (55) and (56) een stukje precedes the non-subject nominal, while in (57) and (58) it follows it. This order is possible in ParCs as well, cf. for example: (i) Je komt hier door [...] de Oesterdam een stukje op te gaan you come here by the oyster-dam a bit-DIM up to go ‘You can get here by going up the Oyster Dam a little’ (http://home.planet.nl/~dijk8929/Duikplaatsen/2a.1.html, May 2008) 46 Http://www.bramvanburg.nl/turftrappers/turftrappers_2007/februari/0207meppel, May 2008.

130 Chapter 6

(58) je longen een stukje op te schonen47 your lungs a bit-DIM up to clean ‘to cleanse your lungs a little’

Constructions (55) and (56) do not specify which part of the process in the base is executed by the trajector (and therefore profiled), just that there is such a part. It could well be that by default we think of an interpretation that starts at one end of the landmark and continues only for a while. The wider context of (55), given here as (59), however, makes it clear that it is a part somewhere in the middle of the process that is executed by the trajector: (59) Flauw, dacht hij, en hij stapte uit, sloot de auto af en liep

ontspannen een stukje de straat door, uit het zicht van eventueel geïnteresseerde buren naar een zijstraat en toen weer terug. (Peper 2003: 374)

‘Petty, he thought, and got out, locked the car and leisurely walked a short way through the street, out of sight of potentially interested neighbors to a side street, and then back again.’

The car is already in the street when the man gets out to begin his little stroll. Nor does his path end at an end of the street: he only goes as far as a side street.

Returning now to example (54), repeated here, I suggest that it can be analyzed in the same way as (55) and (56). (54) De beer loopt snel een stukje de berg op48 the bear walks quickly a bit-DIM the mountain on ‘The bear quickly walks a short way up the mountain’ A ParC with a length modifier evokes the conceptualization of a motion event that satisfies the general constraint (i.e. it goes from not P to completely P), but profiles only part of it.49 In the case of (54) this means that the base consists of the conceptualization of a motion event from the foot of the mountain (not op ‘on’) to where the trajector is completely supported by the mountain (completely op), i.e. maximally to the top. What the construction profiles because of the addition of een stukje is only a portion of this motion event. Which portion is profiled is left unspecified: it 47 Http://irmaschiffers.web-log.nl, May 2008. 48 Http://www.sanmarko.nl/nl/Home-/Vakanties_/Alaska/Homer_-_Anchorage.aspx, May 2008. 49 This analysis could also be applied to examples (51)-(53).


may start at the foot of the mountain, but also halfway up the mountain.50 (54) does strike me as exceptional. It seems to me that the interpretation in which the trajector starts not at (or very close to) a boundary is not available for many ParCs.51 What is required to convey such events is the use of verder ‘further’.

I will finish this section by pointing out that the two types of modification, verder and ‘length modifiers’, may also be combined: (60) Hij kwam overeind he got up

en liep een paar meter verder het strand op and walked a few meters further the beach on en plofte daar weer neer52

and plopped there again down ‘He got up, walked a few meters further onto the beach, and plopped down again there.

(61) dan zien we [...] een jong hert staan [...]. Even later then see we a young deer stand a-little later

loopt het hertje een paar meter verder het bos in om daar weer walks the deer-DIM a few meters further the wood in to there again verder te snuffelen en te knabbelen aan lekkere blaadjes.53

there again further to sniff and to nibble at tasty leaves ‘Then we see a young deer. [...] A little later the little deer is walking a few meters further into the wood to continue sniffing and nibbling at tasty leaves.’

Each construction crucially evokes as part of its base a motion event in which a trajector moves from not P, i.e. a boundary of the landmark, to where it is completely P. The base furthermore includes the information that a trajector has already engaged in this event, and that the subject of the sentence continues to engage in it. The constructions profile the continuation; the length modifiers specify how far the continuation lasts.

As mentioned fn. 35 of this chapter, the first part of the event need not be ‘executed’ by the subject of the ParC: it may be subjective motion (Langacker 2000: Ch. 10) on the part of the language user, i.e. the

50 ParCs with op seem more open to this type of interpretations than other ParCs. 51 Except, of course, for ParCs with uit ‘out’ and af ‘off’, which require that the trajector starts not uit, i.e. IN, or not af ‘off’, i.e. ON, the landmark, cf. section 6.2.3. 52 Http://wobbeenmarieke.waarbenjij.nj/index.php?page=message, May 2008. 53 Http://www.duppen.nl/fotografie/reisverhalen/canada/040509, May 2008.

132 Chapter 6

conceptualizer. The context of (61) makes clear that this is the case here.54 The trajector het hertje ‘the little deer’ is in the landmark het bos ‘the wood’ when the speaker, who is outside of the wood, first sees it. When the deer starts to move, it continues the motion event that the speaker engaged in mentally, i.e. subjectively (Langacker 2000: Ch. 10), to see the deer.55 This subjective motion event goes from the boundary of the wood to where the deer is. The ParC in (61) conveys that the deer continues the motion event by moving from its location in the wood away from the speaker, i.e. away from the boundary of the wood. The constructions discussed in this section may at first sight appear to be counterexamples to the general semantic constraint that I propose for ParCs, because the motion events that they profile do not, or not necessarily, start where the trajector is not P. I have argued, however, that the semantic constraint is imposed in constructions with verder and/or length modifiers as well, but only as part of their base. A ParC with verder profiles the continuation of a motion event that satisfies the general constraint; a ParC with a length modifier restricts the profile to only a portion of the motion event that satisfies the general constraint. 6.6 Semantic differences between ParCs and PreCs The semantic differences between ParCs and PreCs that were identified in this chapter constitute the basis for my analysis of the grammatical constituency of ParCs, which is laid out in the following chapter. We have

54 Cf. the following excerpt: (i) [Dan] zien we dezelfde twee mensen weer als bij de waterval. Ze gebaren

naar het bos en we zien hun mond "deer" zeggen. Zoekend kijken we het bos in en dan zien we inderdaad een jong hert staan op zo'n tien meter van ons vandaan. Hij kijkt ons recht aan. Even blijft hij zo staan, tot hij er zeker van is dat wij geen bedreiging vormen, en dan begint hij aan de blaadjes om hem heen te snuffelen. [...] Even later loopt het hertje een paar meter verder het bos in om daar weer verder te snuffelen en te knabbelen aan lekkere blaadjes. Prachtig! ‘Then we notice the same two people we saw at the waterfall. They are pointing to the wood and mouthing the word ‘deer’. We peer into the wood and notice indeed a young deer at about ten metres away from us. It is looking straight at us. For a moment he remains like that, until he is sure that we do not pose a threat, and then he starts sniffing at the leaves around him. Soon after the little deer walks a few meters further into the wood and continues to sniff and nibble at tasty leaves there. Wonderful!’

55 Incidentally, a ParC describing such a subjective motion event precedes the construction in (61), see the first construction in bold in (i) in fn. 54.


seen that ParCs require traversal, i.e. motion where the landmark is, while PreCs do not. Secondly, ParCs are resultative, while PreCs need not be. And thirdly, ParCs require that the trajector ends up completely P, while PreCs do not.

Chapter 7 The grammatical constituency of ParCs 7.1 Introduction This chapter answers the central question of this study: what is the grammatical constituency of ParCs? It argues that the semantic structure of ParCs can be accounted for in terms of the following grammatical constituency: the adposition and the verb form a complex predicate that takes a subject and a certain type of direct object. The analysis is supported by three semantic parallels. Firstly, just like the particle of a resultative separable complex verb (SCV), the adposition of a ParC is understood as the result of the event, see section 7.2. Secondly, just like direct objects of motion verbs in other languages, the non-subject nominal of a ParC is (partially or completely) traversed, see section 7.3. And thirdly, just like ‘argument/oblique alternations’ (Beavers 2006), pairs of ParCs and PreCs show a ‘holistic/partitive effect’ (Levin 1993: 43-44, 48, and references mentioned there), see section 7.4. Section 7.5 presents a cognitive-grammar account for these observations by proposing that the grammatical constituency of ParCs is not like that of PreCs, but like that of SCVCs. 7.2 Semantic similarities between ParCs and SCVCs The first semantic parallel that I draw is between ParCs and resultative separable complex verb constructions (SCVCs). For ParCs, see (1)-(3), I have argued that the adposition (P) is understood as the result of the event, which means that the event progresses from not P to completely P. (1) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) a girl cycles the Doctor Larij-road on/up ‘A girl is cycling onto the Doctor Larij Road’ (2) hun fietspontje [vaart] het kanaal over (jgdapr95)


(3) [Hij] huppelde […] het Piazza Navona af (de Winter 1989: 305) he skipped the Piazza Navona off ‘He hopped and skipped off the Piazza Navona’

Een meisje ‘a girl’ in (1) moves from where she is not op ‘on’, i.e. from a boundary of de Dokter Larijweg ‘the Doctor Larij Road’, to where she is

136 Chapter 7

completely op, i.e. completely supported by the Doctor Larij Road and no longer by any other surface. Hun fietspontje ‘their bicycle ferry’ in (2) moves so that it is completely over, i.e. from one boundary of het kanaal ‘the canal’ to the opposite boundary. Hij ‘he’ in (3), finally, moves from where he is not af ‘off’, i.e. on het Piazza Navona ‘the Piazza Navona’, to where he is completely af, i.e. no longer supported by the Piazza Navona. The semantic parallel with SCVCs, see (4)-(6), is that in those constructions, too, the adposition is understood as the result of the event (cf. Blom 2005 on SCVs with resultative particles), which means that the event progresses from not P to completely P. (4) Iedereen dronk zijn drankje op (mcmar95ove) everyone drank his drink up ‘Everyone finished their drinks’ (5) de gevangene haalt de trekker over (jnlfeb93) the prisoner pulls the trigger over ‘The prisoner pulls the trigger’ (6) hij [...] maakte de compositie af (mcdec92ove) he made the composition off ‘He finished the composition’ Zijn drankje ‘his drink’ in (4) goes from being not op ‘finished (lit. up)’ to completely op ‘finished’. The result of (5) is that de trekker ‘the trigger’ completely over, that is, it moves from one side of the area in which it is located to the opposite side. De compositie ‘the composition’ in (6) goes from being not af ‘finished (lit. off)’ to completely af ‘finished’. In my analysis, then, ParCs are resultative constructions just like SCVCs such as (4)-(6): the verb and the adposition together specify the event, of which P is understood as the result. A difference between ParCs and SCVCs, however, is that the trajector of the adposition is elaborated by the subject of a ParC and by the direct object of an SCVC. In (1), for example, it is the subject een meisje ‘a girl’ that is completely op ‘on’ as a result of the event, while in (4), it is the direct object zijn drankje ‘his drink’ that is completely op ‘finished’ as a result of the event.

If my analysis is going in the right direction, this means that ParCs constitute a type of resultative construction that Verkuyl (1993) expects not to occur in any language of the world. (7), from Verkuyl (1993: 31), is an ordinary resultative construction: the referents of the direct object mijn voeten ‘my feet’ become warm ‘warm’ as a result of the subject referent’s rubbing them.

The grammatical constituency of ParCs 137

(7) Mirjam wreef mijn voeten warm ‘Mirjam rubbed my feet warm’ Verkuyl (1993: 31) states:

the resultative adjunct [warm] does not modify the subject: Mirjam does not necessarily become warm by rubbing my feet. The only NP that can be modified in [(7)] is the internal argument NP mijn voeten (my feet). There is no a priori reason why the [resultative] adjunct should not modify the external (subject) NP in two-place resultative constructions. Yet this is universally not the case, to my knowledge.

In my analysis, however, ParCs are exactly that: ‘two-place resultative constructions’ in which the subject referent is ‘modified’ by the ‘resultative adjunct’ P. The semantic parallel between ParCs and SCVCs does not stop with the observation that the adposition is understood as the result of the event. In both constructions the subject referent somehow interacts with the referent of the non-subject nominal throughout the event, so that eventually result P is achieved.1 In SCVCs, the interaction consists in the subject referent effecting some kind of change in the direct object referent. The latter participant is therefore referred to by Blom (2005: 124-125) as the affected Theme, which

can be said to measure out the event (Tenny 1992, 1994) and is the incremental Theme in the sense of Dowty (1991): the progression of the event [...] can be measured by looking at this participant [...], the event being partly finished implying this participant to have partly undergone the change of state in question.

Zijn drankje ‘his drink’ in (4) and de compositie ‘the composition’ in (6) being completely finished marks the completion of their respective events, and de trekker ‘the trigger’ being completely ‘on the other side’ marks the completion of the event in (5).

In ParCs, the interaction consists in the subject referent (partially or completely) traversing the direct object referent. The trajector een meisje ‘a girl’ in (1) (partially) traverses the landmark de Dokter Larijweg ‘the Doctor Larij Road’ from one of its boundaries, i.e. where she is not op ‘on’, to where she is completely op. The trajector ons fietspontje ‘our bicycle ferry’ in (2) (completely) traverses the landmark het kanaal ‘the canal’, i.e. from 1 Such interaction is not necessarily required by PreCs, see Chapter 6 and section 7.4 below.

138 Chapter 7

one side of the canal to the opposite side, where the ferry is over. The trajector hij ‘he’ in (3), finally, (partially) traverses the landmark de Piazza Navona ‘the Piazza Navona’ from where he is on (not af ‘off’) the square to where he is completely af ‘off’. The following section zooms in on the role of the referent of the non-subject nominal of ParCs, by considering constructions from other languages in which a motion verb takes a direct object. 7.3 ParCs and motion verbs with direct objects A second semantic parallel that I draw is that between ParCs and constructions from other languages in which a motion verb takes a direct object. Compare, for instance, the Dutch ParC in (2), repeated here, with (8)-(10) from English (Dowty 1991: 569), Japanese (Muehleisen and Imai 1997: 332), and French (Choi-Jonin and Sarda forthcoming). (2) hun fietspontje [vaart] het kanaal over (jgdapr95)


(8) She crossed the desert in a week (9) Jun wa kawa/michi o watatta

Jun TOP river/street ACC cross-PAST ‘Jun crossed the river/street’

(10) Marie a traversé la rivière ‘Marie crossed the river’

Semantically, these constructions are quite similar: a trajector, the subject referent, moves from one boundary of a landmark, i.e. the referent of the non-subject nominal, to an opposite boundary of the landmark. Since the non-subject nominals of (8)-(10) are analyzed as direct objects, het kanaal ‘the canal’ could also very well be a direct object. The non-subject nominal of the Japanese construction takes the accusative case marker -o. Such nominals are known as ‘traversal objects’ (Kuno 1973, Martin 1975, Haig 1981). Kuno (1973: 97) describes the semantics of such constructions as follows: “the motion designated by the verb takes place covering the entire dimension (or the major portion thereof) of the NP continuously and unidirectionally” (1973: 97). Muehleisen and Imai (1997: 335), similarly, observe for such constructions that “there is an implication that a surface is progressively and completely covered by the motion”.


Dowty (1991) analyzes the desert in (8) as a direct object, i.e. as the ‘syntactic realization’ of a type of argument that he calls Incremental Path Theme. Muehleisen and Imai (1997) and Choi-Jonin and Sarda (forthcoming) use the same notion to account for the non-subject nominals in (9) and (10). While I agree that the non-subject nominals are direct objects, I consider the notion of Incremental Path Theme to be problematic, which is an issue that I address first. Dowty’s Incremental Path Theme is a subtype of his Incremental Theme. With Incremental Themes there is a homomorphism between the parts of a telic event and the parts of the Theme, cf. Dowty (1991: 567) on mow the lawn:

If I tell my son to mow the lawn (right now), and then look at the lawn an hour later, I will be able to conclude something about the ‘aspect’ of the event of his mowing the lawn from the state of the lawn, viz., that the event is not yet begun, or is partly done but not finished, or is completed, according to whether the grass on the lawn is all tall, partly short, or all short. By contrast, I will not necessarily be able to inspect the state of my son and conclude anything at all about the completion of his mowing the lawn.

The direct objects in (8)-(10) are considered to be a subtype of Incremental Themes, i.e. Incremental Path Themes, for the following reason. What they share with the lawn is that if the events in (8)-(10) are interrupted before they are completed, it is their direct objects that are ‘partially but not totally affected’ (Dowty 1991: 569).2 In (8), for example, if the event is interrupted, it is the desert that has only been partially, and not completely traversed.

A similar proposal can be found in Tenny (1994). The direct objects in (11) and (12) both, in Tenny’s words, ‘measure out’ the event, which means that they provide “a kind of scale or series of increments that match up with the event at various times or in various stages of completion” (1994: 18). (11) eat an apple (12) Bill climbed the ladder Consider how Tenny describes this for (11) (1994: 15): 2 We may assume that Dowty (1991) is using the term ‘affected’ here purely in the sense of being ‘traversed’. The desert is not affected in the usual sense of undergoing a change of some sort. Tenny (1994) rightly observes for Sue walked the Appalachian Trail and Bill climbed the ladder that the referents of their direct objects “do not necessarily undergo change during the event” (1994: 17-18).

140 Chapter 7

the eating event is understood to progress through the internal argument, the apple, until the end of the apple and of the eating event are achieved. Some quantity of apple is consumed during each interval of eating, until the apple is entirely consumed. In this way the apple provides a measure, in a sense, of the eating event.

In a similar way the event of climbing the ladder in (12) is measured out by the length of the ladder: the end of the event is defined by the end of the ladder.

Tenny is criticized by Dowty, however, for proposing that ‘measuring out’ is an exclusive property of direct objects. According to Dowty, subjects can be Incremental Path Themes too, cf. (13)-(15) from Dowty (1991: 570-571): (13) At the turtle race, the winning turtle crossed the finish line in 42

seconds. (14) John entered the icy water (very slowly). (15) The crowd exited the auditorium (in 21 minutes). Dowty argues as follows. The direct object the finish line in (13) cannot be an Incremental Path Theme because it is a line and not a region. The subject referent the winning turtle, however, is a region that traverses the finish line gradually and is therefore an Incremental Theme. For (14) and (15), Dowty (1991: 571) proposes that the meanings of enter and exit

treat the stationary threshold or boundary traversed (and the direct object referent) as a line or plane rather than a region, but allow that a space-occupying body (the subject referent) traverses it gradually, which means the subject is an Incremental Theme.

What Dowty preserves, in other words, in his analysis of (13)-(15) is that there is complete traversal: the finish line in (13) is completely traversed, as are the icy water in (14) and the auditorium in (15), which are ‘treated’ as lines or planes. Yet, these referents do not provide increments that, in Tenny’s (1994: 18) words, “match up with the event [...] in various stages of completion”. These increments are provided by the subjects. As I see it, there are two problems here. The first is that it seems counter-intuitive to propose that the direct objects in (14) and (15) are understood as ‘stationary threshold or boundary’: do not the icy water and the auditorium refer to the entities as wholes, rather than just one of their boundaries or surfaces? Consider, for example, Langacker’s (1987: 244) characterization of the semantic structure of enter:


The essential import of [ENTER] is that a trajector, through time, progresses from an [OUT]-relation to an [IN]-relation with respect to some landmark.

His notions [OUT]-relation and [IN]-relation presuppose that the landmark is a container of some sort, not a ‘stationary threshold or boundary’. The second problem is that it seems equally counter-intuitive to analyze the direct objects of cross in (8) and (13) in such different ways: the desert in (8) is an Incremental Path Theme, but the finish line in (13) is not.

A more unified analysis of these examples is possible, I suggest, once it is realized that there is an essential difference between mow the lawn or eat an apple, on the one hand, and cross the desert or exit the auditorium, on the other hand. In the case of mow the lawn and eat an apple, it suffices to inspect only the lawn or only the apple to draw conclusions about the progression of the event. This is not so in the case of examples such as cross the desert. Inspecting only the desert does not suffice to draw conclusions about the progression of the event. Consider Fig. 1, which represents the desert at three different moments of the event: only if we also take into account the position of the subject referent can we draw conclusions about whether the event has just started, is halfway, or completely finished, see Fig. 2, which includes a representation of the trajector.

142 Chapter 7

Similarly in (13), just considering the subject referent does not

suffice for drawing conclusions about the progression of the event, see Fig. 3, which represents the winning turtle at different times of the event. (13) At the turtle race, the winning turtle crossed the finish line in 42

seconds. The subject referent and object referent have to be considered both to draw such conclusions, see Fig. 4, which represents both the winning turtle and the finish line.

t

Fig. 1. The desert at different times of the event

t

Fig. 2. She and the desert at different times of the event


The same can be observed for climb the ladder in (12), enter the icy water in (14), and exit the auditorium in (15): only if both the subject referent and the referent of the direct object are taken into account, can we draw conclusions about the progression of these different events, see e.g. Figs. 5 and 6 representing the events of (14) and (15).

Fig.3. The winning turtle at different times of the event

t

Fig. 4. The winning turtle and the finish line at different times of the event

t

Fig. 5. John and the icy water at different times of the event

Fig. 6. The crowd and the auditorium at different times of the event

144 Chapter 7

These constructions are different from mow the lawn or eat an apple because there is not one participant that ‘measures out’ the event. This is inherent in the type of event involved: the events designated by cross, enter, and exit consist in a trajector moving with respect to a landmark. To judge whether the event has been completed, one has to look at the trajector’s position with respect to the landmark. Two questions now present themselves. Firstly, do the direct objects of (8)-(10) and (12)-(15) have something in common semantically with the direct objects of mow the lawn and eat an apple? And secondly, what constitute the beginning and endpoints of such events?

Dowty’s ‘completely affected’ Incremental Path Theme or Tenny’s ‘measuring out’ cannot be the semantic generalization for these direct objects, because with enter and exit the referent of the direct object is only partially traversed. What they do seem to have in common with more prototypical direct objects is that the event progresses through the direct object referents, cf. Tenny on eat an apple: “the eating event is understood to progress through the internal argument, the apple” (1994: 15). There is no requirement, however, that the direct object is ‘progressed through’ completely, i.e. from one of its boundaries to an opposite boundary.

What marks the beginning and endpoint of an event designated by enter is similar to that of ParCs with in, compare e.g. (14) with (16). (14) John entered the icy water (very slowly). (16) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’ They each designate an event that progresses from not in to completely in, cf. Langacker’s characterization of the semantics of enter quoted above:3 the trajector (the subject referent) moves from where it is not in, i.e. a boundary of the landmark (the direct object referent), to where it is completely in, that is no longer contained by anything else.4

3 There does seem to be a semantic difference between the Dutch and English constructions. The events designated by enter and exit appear to be finished when the trajector is completely in or out of the landmark, i.e. when the trajector has crossed one of the landmark’s boundaries completely. The events designated by Dutch ParCs with in or uit ‘out’ can be more extensive: they can continue as long as the trajector is construed as heading towards the result in respectively uit; the length of the path may be specified by a ‘length modifier’. 4 The constraint that the trajector is completely in the landmark is satisfied also when, for example, John’s head is still sticking out from the water: he is completely


Compare also (15) and the ParC with uit ‘out’ in (17). (15) The crowd exited the auditorium (in 21 minutes). (17) Ze lopen […] het dorp uit (Enquist 1996: 80)

they walk […] the village out ‘They are walking out of the village’

They each designate an event that progresses from not OUT to completely OUT, i.e. from where the trajector is in the landmark to where it is completely out of the landmark.

The events in (14)-(17) (partially) ‘progress through’ the referent of the direct object: the trajector’s motion takes place where the landmark is. This is, I propose, what ParCs and the English construction with motion verbs and direct objects share with affected direct objects such as an apple in eat an apple. ‘Progression through’ is an abstraction over affected direct objects and how I redefined ‘traversal’ in 6.2, i.e. as motion only where the landmark is. What Dutch ParCs share with English constructions with cross, enter, and exit is that their events progress through the landmark from where the trajector is not ACROSS, IN, or OUT OF the landmark to where it is completely ACROSS, IN, or OUT OF the landmark. 7.4 ParCs, PreCs, and the holistic/partitive effect The third semantic parallel, finally, revolves around the holistic/partitive effect. This section argues that pairs of ParCs and PreCs show this effect just like ‘argument/oblique alternations’, in which “a single argument can be realized as either a direct argument (subject/object) or an oblique” (Beavers 2006: v). This is a third argument for an analysis of Dutch ParCs in terms of a complex predicate with a subject and a direct object: the holistic/partitive effect in ParCs and PreCs can be accounted for, if we assume that the non-subject nominal is a complement of a preposition in the PreC, but a direct object in the ParC.

The prime example of an argument/oblique alternation showing the holistic/partitive effect is the locative alternation, see (18) and (19) from Levin (1993: 51). (18) Jack sprayed the wall with paint (19) Jack sprayed paint on the wall

contained in the water in the sense that he is no longer supported by the land surrounding the water.

146 Chapter 7

The holistic/partitive effect can be observed for the wall, which is a direct object in (18) and a complement of a preposition in (19). A semantic difference between (18) and (19) is that the wall is completely covered (‘totally affected’, Hopper and Thompson 1980) by paint as a result of the event in (18), but not necessarily so in (19) (just ‘affected’).

A second example is the conative alternation illustrated by (20) and (21) from Beavers (2006: 64). (20) The forlorn diner ate his sandwich. (21) The forlorn diner ate at his sandwich. His sandwich is a direct object in (20) and is eaten completely, while as a complement of the preposition in (21) it need be eaten only partially. The ‘holistic/partitive’ effect is, in other words, observed for the entity that is expressed as a ‘bare nominal’ in one construction and as the complement of a preposition in the other. The direct objects in (18) and (20) are completely affected by the event, while the complements of on and at in (19) and (21) need only be partially affected by the event.

A final example, and one particularly relevant to the present study, is the so-called ‘preposition drop alternation’ illustrated by (22) and (23) from Beavers (2006: 69). (22) The hiker walked the Barton Springs Trail. (23) The hiker walked up/along/down the Barton Springs Trail. (22) is a construction with a motion verb that takes a direct object, the Barton Springs Trail. The implication is that the Barton Springs Trail is ‘completely traversed’ in (22). In (23), where it is the complement of a preposition, it need not be completely traversed: partial traversal is all that is required.

The semantic contrast between (22) and (23) is very similar to the semantic contrast between a ParC and PreC such as (2), repeated here, and (24): (2) hun fietspontje [vaart] het kanaal over (jgdapr95)


(24) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrockcanal ‘We are sailing along the Haren-Rüttenbrock Canal’


The non-subject nominal in the ParC in (2), het kanaal ‘the canal’, is completely traversed: the trajector (het fietspontje ‘the bicycle ferry’) moves from one boundary of the canal to the opposite boundary of the canal. The non-subject nominal in the PreC in (24), het Haren-Rüttenbrockkanaal ‘the Haren-Rüttenbrock Canal’, on the other hand, is just partially traversed: the trajector (we) follows a path along the surface of the canal, without any further specifications with respect to its length. It would be incorrect, however, to characterize the holistic/partitive effect in ParCs and PreCs in terms of ‘complete traversal’ vs. ‘partial traversal’ contrast. Complete traversal is required by ParCs with path adpositions such as over ‘over’, door ‘through’, and langs ‘along’, but not by ParCs with point adpositions such as op ‘on’, in ‘in’, binnen ‘inside’, af ‘off’, and uit ‘out’. The latter type of ParCs requires only partial traversal: (25) Johan Cruijff sprintte het veld op (jgdmay92)


(16) T. reed [...] de Leliestraat in (mcoct92spo) T. drove [...] the Lily-street in ‘T. drove into Lily Street’ (26) Marconi ging de kamer binnen (stdec1995) Marconi went the room inside ‘Marconi went into the room’ (3) [Hij] huppelde […] het Piazza Navona af (de Winter 1989: 305)

he skipped the Piazza Navona off ‘He hopped and skipped off the Piazza Navona’


Johan Cruijff in (25), for example, partially traverses het veld ‘the field’ from one of its boundaries to where he is completely op, i.e. completely supported by the field. Ze ‘they’ in (17), similarly, partially traverse het dorp ‘the village’ from a position where they are not uit ‘out’, i.e. in the village, to a boundary of the village, i.e. where they are completely uit. Notice, however, that the ‘holistic’ notion completely does figure in the semantic descriptions of these ParCs, which is why I suggest that ParCs and PreCs do show a holistic/partitive effect. It is not found in their non-

148 Chapter 7

subject nominals, however, but in their subjects.5 Regardless of whether a ParC contains a point or a path adposition, it imposes the constraint that the trajector ends up completely P. As a result of the motion event in (2), for example, the subject referent hun fietspontje ‘their bicycle ferry’ is completely over, i.e. on the other side of the landmark het kanaal ‘the canal’. Similarly, as a result of the motion event in (25), the subject referent Johan Cruijff is completely op, i.e. completely supported by het veld ‘the field’.

PreCs do not impose such a constraint. The subject referent of the PreC in (24), we, need not be completely over at the end of the event, nor need the subject referents of the PreCs in (27) or (28) be completely op. (24) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrockcanal ‘We are sailing along the Haren-Rüttenbrock Canal’ (27) Het legervoertuig […] reed […] op een mijn (jgdoct93) ‘The army vehicle drove on a mine’ (28) Op 3 december [...] voer een schip [...] op de rotsen bij La Coruña (jgddec92) on 3 December sailed a ship on the rocks near La Coruña ‘On December 3 a ship ran (lit. sailed) on the rocks near La Coruña’ The landmark in (27), een mijn ‘a mine’, is too small to support het legervoertuig ‘the army vehicle’ completely, see Fig. 7.6 The trajector in (28), een schip ‘a ship’ probably stops moving once it hits the landmark de rotsen ‘the rocks’, so that it ends up being in contact with the rocks only partially, see also section 6.3.

5 Cf. also Beavers (2006: 152) on the preposition drop alternation in English: “the relevant lexical entailments are not associated with the alternating participant, but rather with the figure participant, as shown [...] in [(i) and (ii)] (i) The hiker climbed the mountain. (ii) The hiker climbed up the mountain. Here it is the figure the hiker that is entailed to have been totally affected in [(i)] (the hiker goes from a specific source state to a specific goal state) and to have been affected in the [(ii)] (going from a non-specified source state to a non-specified goal state)”. 6 I am assuming a default context here, with default sizes for het legervoertuig ‘the army vehicle’ and een mijn ‘a mine’. Cf. also the discussion of Jan stapt op de kiezelsteen ‘John steps on the pebble’ in section 6.3.


In summary, ParCs and PreCs show a holistic/partitive effect that can be observed in their subjects: the subject referent of a ParC is completely P as a result of the motion event, while this is not required in the case of PreCs (their subject referents may end up being just partially P). 7.5 ParCs and PreCs: different grammatical constituency 7.5.1 ParCs and PreCs: different semantics What ParCs and PreCs have in common semantically is that they profile motion events in the sense of Talmy (1985: 60-61): a trajector moves with respect to a landmark. The points on which they semantically differ, on the other hand, are exactly the points of similarity with the ‘non-oblique constructions’ discussed in sections 7.2-7.4, which leads me to the conclusion that ParCs are ‘non-oblique constructions’ too. This subsection recapitulates the semantic parallels that have been drawn in the previous sections.

For ParCs I have proposed the semantic constraint described in (29): (29) ParC profiles a motion event in which a trajector traverses a

landmark so that result P is achieved: the trajector moves from where it is not P to where it is completely P.

ParCs with op, such as (1) or (25), for example, can be represented as Fig. 8, in which the circle represents the trajector, the rectangular represents the landmark, and the arrow represents the trajector’s motion.

Fig. 7. Trajector partially supported by landmark

150 Chapter 7

(1) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) a girl cycles the Doctor Larij-road on/up ‘A girl is cycling onto the Doctor Larij Road’ (25) Johan Cruijff sprintte het veld op (jgdmay92)


PreCs, conversely, are not necessarily resultative. In (30), for

example, the subject referent’s motion takes place op de Prinsenstraat ‘on Prince Street’: the PP is a modifier whose trajector is elaborated by de man reed ‘the man was driving’, see section 5.4.2. (30) De man reed op de Prinsenstraat (mcfeb93ove) ‘The man was driving on Prince Street’ This motion event can be represented as in Fig. 9a: the circle represents de man ‘the man’ and the rectangular represents de Prinsenstraat ‘Prince Street’. I have not represented the motion that de man ‘the man’ engages in.

Secondly, PreCs do not require traversal. (31), for example, is resultative: the trajector ends up being op de grond ‘on the ground’ as a result of the motion designated by the verb sprong ‘jumped’. (31) Japie [sprong] op de grond (gp94-2) ‘Japie jumped on the ground’ (31) allows an interpretation in which the trajector Japie jumps straight down, from a tree, for example. This is represented in Fig. 9b, in which the circle represents the trajector, the horizontal line represents the landmark, and the vertical arrow represents the motion. I have defined traversal as ‘motion where the landmark is’, which in ParCs with op means that the trajector covers some part of the surface of the landmark while it moves, if not in contact with the landmark than close to it. This is not required by the PreC in (31).

Fig. 8. ParC with op


And thirdly, PreCs do not require that the trajector ends up being completely P, which was illustrated by means of (27) and (28) above. This can be represented schematically as I have done in Fig. 9c: only part of the trajector is supported by the landmark; it is also supported by some other surface, represented by the thin horizontal line.

ParCs and PreCs differ in these three respects, resultativity,

traversal, and completely P, and it is exactly those respects in which Parcs are similar to the ‘non-oblique constructions’ discussed in sections 7.2-7.4. ParCs share with SCVCs such as (4) that (i) the adposition is understood as the result of the event, (ii) that the event progresses from not P to completely P, and (iii) that the event ‘progresses through’ the direct object. (4) Iedereen dronk zijn drankje op (mcmar95ove) everyone drank his drink up ‘Everyone finished their drinks’ The latter two aspects are also present in constructions from other languages in which a motion verb takes a direct object, such as (8)-(10), from English, Japanese, and French. (8) She crossed the desert in a week (9) Jun wa kawa/michi o watatta

Jun TOP river/street ACC cross-PAST ‘Jun crossed the river/street’

(10) Marie a traversé la rivière ‘Marie crossed the river’

The requirement imposed by ParCs that the trajector ends up being completely P, while PreCs require just that the trajector ends up being partially P, finally, was argued to be a type of holistic/partitive effect.

Fig. 9a. Not resultative

Fig. 9b. No traversal

Fig. 9c. Trajector not completely P

152 Chapter 7

As stated at the beginning of this section, the semantic similarities between ParCs and ‘non-oblique constructions’, on the one hand, and the semantic differences between ParCs and (oblique) PreCs, on the other hand, lead me to conclude that ParCs are ‘non-oblique constructions’ too. I propose that the semantic differences between ParCs and PreCs can be accounted for in terms of different hierarchical ways of combining the semantic structures of the words involved. A crucial difference is that in a ParC, the adposition combines with the verb, while in a PreC, the adposition combines with the nominal that follows it, which has a different semantic effect. The remainder of this section gives a cognitive-grammar sketch of how I envisage this for ParCs and PreCs with op and over 7.5.2 Grammatical constituency in Cognitive Grammar As mentioned in 5.4.1, Cognitive Grammar does not distinguish syntactic structure as a separate level of linguistic organization. Grammar consists only of phonological structures, semantic structures, and symbolic structures, which are pairings of phonological structure and semantic structure. Grammatical constituency in this framework is, as Langacker (1995: 150) puts it, “recognized but is not accorded autonomous status: it is merely the order in which simpler symbolic structures combine to form progressively larger ones”. Consider, for example, the prepositional-phrase construction in Fig. 10, which was discussed in 5.4.1. The representation shows that the symbolic structure in integrates with the symbolic structure the car to form the composite symbolic structure in the car. Langacker (2000: 379) states the following about representations such as Fig. 10:

These assemblies are quite different from the tree structures of generative grammar, which are generally conceived as autonomous syntactic objects, devoid of intrinsic semantic or phonological content. Observe that each ‘node’ in Figure [10] is a symbolic structure with both a semantic and a phonological pole. At a given level of organization, grammatical class is determined by the composite structure’s profile [...], in contrast to the ‘syntactic’ (hence semantically empty) node labels of generative phrase trees. Another difference is that the symbolic elements in [Fig. 10] are not linearly ordered with respect to one another (i.e. the hierarchical structure is more like a mobile than a tree with fixed branches). So-called ‘linear’ order is actually temporal order, which each symbolic element specifies internally as part of its phonological characterization.


Fig. 10. A representation of the grammatical construction in the car

In the car is an example of what Langacker (1997) calls a ‘classical grammatical constituent’, i.e. a symbolic pairing of a ‘classical conceptual constituent’ and a ‘classical phonological constituent’. The semantic structure [IN THE CAR] forms a ‘classical conceptual constituent’, in that “two component structures combine semantically on the basis of [...] a correspondence between salient substructures” (1997: 13), i.e. the profile of [THE CAR] corresponds to the landmark of [IN]. The phonological structure [in the car] forms a ‘classical phonological constituent’, in that [in] and [the car] are “two component structures [that] form a group on the basis of temporal contiguity” (1997: 13). The classical conceptual constituent [IN THE CAR] and the classical phonological constituent [in the car] form the classical grammatical constituent in the car.

The grammatical relations of subject and direct object are not defined configurationally, but semantically. Cognitive Grammar provides a characterization of prototypical subjects (agent, human, definite, figure within the profiled relationship) and direct objects (patient, physical object, specific indefinite, secondary figure, see Langacker 1991a: Ch. 7). In addition, it provides schematic characterizations, i.e. what is common to all

the car

in the car

lm

tr

in

tr

lm

154 Chapter 7

subjects and what is common to all direct objects, see Langacker (1995: 154):

A direct object is [...] a noun phrase whose profile corresponds to the landmark of the process profiled by the clause as a whole. [...] A subject is a noun phrase whose profile corresponds to the trajector of the process profiled at the clausal level of organization.

Verbs designate ‘processes’, which, just like other relational

predicates such as adpositions, are conceptually dependent, while nominals profile ‘things’, i.e. regions in some domain (matrix), which are conceptually autonomous, see Langacker (1991a: 286):

An event is conceptually dependent vis-à-vis its participants. For instance, one cannot conceptualize an act of slapping without making some kind of mental reference to the entity doing the slapping and the one receiving it – however vaguely these entities might be portrayed, if they are absent altogether the conception is incoherent. The verb slap is thus described as referring schematically to these participants as salient facets of its semantic structure. By contrast, a person or a physical object can be conceptualized independently of any event in which it might participate; although it is part of what we know about such entities that they do function as event participants, our conception of them is coherent even if this knowledge remains latent. They are, in short, conceptually autonomous.

Just as op relates two entities, so does the verb slap. There is an asymmetry between the two entities: one is the primary figure, i.e. the trajector, the other is the secondary figure, i.e. the landmark. In a clause such as She slapped him, the profile of she corresponds to the trajector of slapped, which makes she the subject of the clause; the profile of him corresponds to the landmark of slapped, which makes him the direct object of the clause. 7.5.3 The grammatical constituency of ParCs and PreCs The semantic differences between ParCs and PreCs can be related, I propose, to their different grammatical constituency, i.e. the different hierarchical ways in which their components combine. In a ParC, the adposition combines with the verb to form a complex predicate which necessarily requires two participants: a trajector and a landmark. In a PreC, the adposition combines with the nominal that follows it to form a prepositional phrase; the prepositional phrase combines with the verb, which requires just one participant: a trajector. The details of this proposal, including the


semantic effects of the different constituency, are explained in this subsection, which focuses first on PreCs, and then on ParCs. Conceptual integration between two components can take place if the components have, or can be construed as having, substructures that can be put in correspondence. A prepositional phrase, as we have seen, consists of two components: an adposition and a nominal. They can integrate by virtue of their having substructures in common: the nominal elaborates, i.e. specifies in more detail, the landmark that is part of the semantic structure of the adposition.

The elaboration of the landmark makes the relation designated by the adposition more specific. For over, for example, Chapter 4 proposes the following semantic constraint: (32) Over designates a relation between a trajector and a landmark in


In the PreC in (24), over combines with the nominal het Haren-Rüttenbrockkanaal. (24) We varen over het Haren-Rüttenbrockkanaal (wk199212) we sail over the Haren-Rüttenbrockcanal ‘We are sailing along the Haren-Rüttenbrock Canal’ The profile of het Haren-Rüttenbrockkanaal ‘the Haren-Rüttenbrock Canal’ elaborates the landmark in the semantic structure of over, which makes the relation more specific. Over het Haren-Rüttenbrockkanaal designates a relation between a trajector and the canal in which the trajector is related to the canal by a mental path that follows a surface of the canal, and from which a force points to the canal. The elaboration of the landmark has a specific semantic effect, i.e. it narrows down the interpretative possibilities in a particular way. The prePP in (24) specifies that only a portion of the canal is ‘involved’ with the trajector, namely that portion of the landmark that is ‘covered’ by the path.

Fig. 11. Dutch over

156 Chapter 7

Fig. 11 shows that the constraint imposed by over involves a path that follows the surface of the landmark; the constraint does not require a path that completely traverses the landmark. This is, in my view, the explanation of the ‘partitive effect’ in constructions with PPs. In PreCs, the prePP combines with a verb of motion. Such verbs can be characterized in the way that Langacker (1987: 305) does for crawl:7

Crawl combines two kinds of relational specifications: (1) the schematic path followed by its trajector with respect to external surroundings; and (2) the activity the trajector carries out to propel itself through this spatial trajectory.

The semantic structure of the prePP in (24) integrates with the verb varen ‘sail’: the path in the semantic structure of the PP elaborates the schematic path in the semantic structure of the motion verb, and the trajector of the PP corresponds to the trajector of the verb. The trajector is elaborated by the subject of (24), we ‘we’. The ‘partitive effect’ is that the trajector we follows a path with respect to the surface of the landmark, one that does not necessarily traverse it.

With point adpositions the necessary ‘involvement’ of the landmark in the event is even more minimal: it is limited to a point. To show this, I state what I consider to be the semantic constraint imposed by Dutch op in (33), and I represent it visually in Fig. 12:8 (33) Op profiles a relation between a trajector and a landmark in which

the trajector is in contact with a surface of the landmark, and there is a force that points from the trajector to the landmark.

7 Crawl is a manner of motion verb. Most verbs that occur in ParCs are manner of motion verbs, but other types of motion verbs are possible too, such as komen ‘come’ and gaan ‘go’. The second kind of relational specification in Langacker’s characterization does not apply to those verbs. 8 See for the differences between Dutch op and aan, which both translate as ‘on’, Cuyckens (1991), Bowerman (1996a), Beliën (2002), and van Staden et al. (2006). A potential problem for the monosemous analysis for op that I propose here is that it does not seem to capture what is going on in cases such as (i) and (ii), in which the trajector is not in contact with a landmark. (i) De rook steeg op the smoke rose up ‘The smoke rose’ (ii) Hij stond/keek op ‘He got/looked up’


The circle in Fig. 12 represents the trajector, the rectangular represents the landmark, and the arrow represents the force.

Before I go into the semantic effect of the elaboration of op’s landmark, I want to note that Fig. 12 should not obscure the following three points. First of all, the constraint for op is neutral with respect to whether the trajector is higher than, lower than, or to the side of the landmark, cf. my proposal for over in Chapter 4. (33) specifies that the trajector is in contact with the landmark and that there is a force going from the trajector to the landmark. As long as there is such a force and contact, the trajector can be higher than the landmark, to the side of it or below it, see (34): (34) Er zit een vlieg op de vloer/op de muur/op het plafond ‘There is a fly on the floor/on the wall/on the ceiling’ Secondly, the constraint does not specify the extent of the contact between the trajector and the landmark. Contact minimally requires point-like contact, as in (35), but more extensive contact, as in (36), satisfies the constraint as well. (35) Er zat een minuscuul stofje op de lens ‘There was a minuscule speck of dust on the lens’ (36) Er lag [...] een zeer vuil tapijt op de grond9 ‘There was a very dirty carpet on the ground’ And thirdly, while Fig. 12 represents the trajector as being completely supported by the landmark, the constraint is neutral in that respect too. I will come back to this point below. When op combines with a nominal to form a prePP, the semantic effect is that what is specified is that a point of the surface of the elaborated landmark is necessarily involved with the trajector. In (30), for example, the prePP op de Prinsenstraat ‘on Prince Street’ designates a relation between a trajector and Prince Street in which the trajector is in contact with a surface

9 Http://www.zoover.nl/reizen?N=29+70, November 2007.

Fig. 12. Dutch op

158 Chapter 7

of Prince Street, and there is a force that points from the trajector to Prince Street. (30) De man reed op de Prinsenstraat (mcfeb93ove) ‘The man was driving on Prince Street’ The prePP is a modifier, see section 5.4, i.e. it integrates conceptually with the verb as follows: the trajector of the PP is elaborated by the process designated by reed ‘drove/was driving’. That PreCs with point adpositions do not require traversal can now also be explained, see e.g. (31). (31) Japie [sprong] op de grond (gp94-2) ‘Japie jumped on the ground’ Conceptual integration of op and de grond ‘the ground’ in (31) results in a relation in which an unelaborated trajector is in contact with the ground’s surface. It specifies, in other words, that there is minimally a point of contact between the trajector and the ground. The prePP is a complement: its semantic structure elaborates a substructure of the motion verb sprong ‘jumped’. This requires a ‘focal adjustment’ (cf. Langacker 1987: 305 and fn. 15 of section 5.4.2 above): the path inherent in the semantics of sprong ‘jumped’ is construed as having an endpoint. The point in the semantics of op de grond ‘on the ground’ elaborates the endpoint of sprong ‘jumped’:10 the prePP’s trajector corresponds with the trajector of sprong ‘jumped’.

A crucial point here is that the landmark de grond ‘the ground’ need not be traversed by the trajector. The prePP specifies that there must be a point of contact, which is construed as the endpoint of the trajector’s motion. The schematic path followed by the trajector inherent in the semantics of sprong ‘jumped’ takes place ‘with respect to external surroundings’ (Langacker 1987: 305) that are not specified any further.11 The motion that

10 Or, alternatively, the semantic structure of the verb is construed as resulting in a spatial relation that corresponds to the spatial relation designated by the prepositional phrase. 11 The ‘external surroundings’ are specified at a high level of schematicity and do not require further elaboration. The semantic structure of zwemmen ‘swim’, for example, includes knowledge that the motion takes place in water, just as the semantic structure of vliegen ‘fly’ includes knowledge that the motion takes place in the air. These are not, however, ‘open places’ in the semantics of the verbs that require further specification before any coherent conceptualization can arise: the schematic information about the external surroundings provided by the verb itself


the trajector engages in ‘only’ ends in the contact relation designated by prepositional phrase: there is no specification that the path needs to traverse the landmark inherent in the semantics of the adposition. Only at the end of the jump is the trajector in contact with the landmark.

Consider in this respect van der Leek’s (1996) explanation for the holistic/partitive effect observed for examples such as (20) and (21), repeated here: (20) The forlorn diner ate his sandwich. (21) The forlorn diner ate at his sandwich. Van der Leek (1996: 371) explains the effect by pointing to the semantics of the prepositional phrase:

the at-phrase, by merely asserting a point of contact between the ingester’s mouth/teeth etc. and the entity subjected to ingestion, explicitly leaves it open how much of this entity is, in fact, ingested.

As I said, the constraint imposed by op ‘on’ is neutral with respect to partial or complete support: the trajector is in contact with the landmark and there is a force that points from the trajector to the landmark. Fig. 6 shows for (27) that the trajector, het legervoertuig ‘the army vehicle’, is in contact with the landmark, een mijn ‘a mine’, and that there is a force that goes from the trajector to the landmark: the part of the vehicle that is in contact with the mine, e.g. one of its wheels, is pulled towards the mine by gravity. (27) Het legervoertuig […] reed […] op een mijn (jgdoct93) ‘The army vehicle drove on a mine’

needs no further elaboration. The schematic trajector, conversely, is such an open place: these ‘manner of motion’ processes are dependent on the conceptualization of a person or animal that carries them out.

Fig. 6. Trajector partially supported by landmark

160 Chapter 7

Similarly, op de rotsen ‘on the rocks’ in (28) specifies that a trajector is in contact with de rotsen ‘the rocks’, i.e. that a trajector is in contact with, minimally, a point of the rocks. (28) Op 3 december [...] voer een schip [...] op de rotsen bij La Coruña (jgddec92) on 3 December sailed a ship on the rocks near La Coruña ‘On December 3 a ship ran (lit. sailed) on the rocks near La Coruña’ The prePP is a complement; it elaborates the endpoint of the spatial path in the semantics of voer ‘sailed’. The motion need not, in other words, traverse the rocks: only at the end of the path is the trajector in contact with the rocks. Nor need the trajector be completely on the rocks: as long as it is in contact with it, and there is a force going from the trajector to the landmark, the constraint is satisfied. Let us now turn to ParCs, whose semantic structure does involve resultativity, traversal, and completely P. These aspects can be accounted for, I propose, if we assume that the construction consists of a complex predicate (P+V). That is, the semantic structure of the adposition (P) is integrated with the semantic structure of the motion verb so that together they designate an event that has P as its result.12

While a manner of motion verb ‘on its own’ requires just one participant, the complex predicate P+V is conceptually dependent on two participants. The reason for this is that P, the result of the event, designates a spatial relation between two entities: a trajector and a landmark. The trajector of P corresponds to the trajector of the motion verb. For the trajector to achieve the result, a landmark is also necessarily involved in the event.

Consider, for instance, the ParC in (1). (1) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) a girl cycles the Doctor Larij-road on/up ‘A girl is cycling onto the Doctor Larij Road’ Fietst ‘cycles’ as such is conceptually dependent on just one participant, cf. Langacker’s characterization of crawl quoted earlier. But in (1) it combines

12 When the phonological structure of the verb is not preceded directly by the phonological structure of the adposition, as in (1), they do not form a ‘classical grammatical constituent’ in the sense of Langacker (1997: 13). Conceptually, however, they do form a unit, on the basis of which we can call the adposition and the verb a constituent, be it not one of the classical kind.


with op to form a complex predicate. Just as in a SCVC, op ‘on’ is understood as the result of the event, and the event progresses from not op to completely op. The non-subject nominal de Dokter Larijweg ‘the Doctor Larij Road’ is a direct object: its profile elaborates the landmark of the event. Just as in other direct object constructions, the event ‘progresses through’ this direct object, be it only partially: the trajector interacts with the landmark throughout the event. The trajector starts at a boundary of the landmark, where it is not op, and (partially) traverses the landmark to where it is completely op. Let us now consider the integration of a verb of motion and a path adposition, see e.g. (2). (2) hun fietspontje [vaart] het kanaal over (jgdapr95)

their bicycle-ferry sails the canal over/across ‘Their bicycle ferry sails across the canal’

The language user has to find a conceptualization that satisfies the constraint imposed by over and the constructional constraint that over is a result. The only way to satisfy this, in a motion scenario, is that the trajector is, calculated from a contextually recoverable vantagepoint, on the other side of the landmark. Such a spatial configuration satisfies both constraints: the mental path that the conceptualizer scans satisfies the constraint imposed by over; the trajector’s position on the other side satisfies the constraint that a result has been achieved, i.e. the trajector has completed a path from one end of the canal to the opposite side. Further support for the idea that P and V form a complex predicate that is conceptually dependent on two participants comes from examples such as (37) and (38), which impose the same constraints as ParCs, but ‘lack’ the nominal elaborating the landmark.13 (37) zodat de pont kan overvaren (wk199318) so that the ferry can over-sail ‘so that the ferry can sail across’

13 Not mentioning the non-subject nominal is not possible for all P+V combinations, see for instance: (i) Hij ging de kamer in/*Hij ging in

he went the room in ‘He went into the room’

(ii) Hij wandelde de berg op/*Hij wandelde op he hiked the mountain up ‘He hiked up the mountain’

162 Chapter 7

(38) Zodra de chauffeur is ingestapt (jgdoct93) as-soon-as the driver is in-stepped ‘As soon as the driver has gotten in’ In (37), there needs to be complete traversal, from one boundary of the implicit landmark to the opposite boundary (we assume that there is some body of water that is crossed by the ferry). In (38), the driver needs to end up completely in the landmark (we assume a vehicle of some sort) (as opposed to PreCs in which it is possible for the trajector to be only partly contained in the landmark, while still supported for the larger part by some other surface). The different grammatical constituency of ParCs and PreCs and the associated semantic effects can be summarized as follows. In the case of a ParC, P and V specify the motion event together: the motion is, all the while, i.e. throughout the event, oriented towards result P. P designates a relation between a trajector and a landmark. The trajector is elaborated by the subject, which is also the trajector of the motion. The subject’s motion is oriented towards result P. This requires another entity, one that corresponds to the schematic landmark of P, with respect to which the subject moves. The motion event develops vis-à-vis the landmark: it starts where the trajector is not P relative to the landmark, and it ends where the trajector is completely P relative to the landmark; all the while, the trajector moves only where the landmark is. In the case of a PreC, the verb of motion requires just one participant, a trajector: it moves in the way that the verb describes. The conceptualization of a landmark is not required by the verb. P and the non-subject nominal form a composite structure together, a prepositional phrase (prePP), which designates a spatial relation between a (schematic) trajector and an (elaborated) landmark. Because the landmark is specified and because we know the meaning of P, we come up with a composite semantic structure in which a portion of space is singled out with respect to the landmark: point-like in the case of ‘point’ adpositions vs. path-like in the case of ‘path’ adpositions.14 We know that the trajector that is schematically

14 This semantic difference between Dutch ParCs and PreCs, i.e. completely P vs. partially P can also be observed in the examples that Capelle (2004) provides to support his claim that English particles and prepositional phrases are not always semantically equivalent, cf. (i)-(ii) from Capelle (2004: 51). (i) [The supersonic aircraft Concorde will make its first test flight from

London on Tuesday since the fleet was grounded following a crash in Paris last year, which killed 113 people.] British Airways has confirmed that the plane will fly over the Atlantic for nearly three and a half hours before returning to its base.


present in the composite structure prePP is located or moving at this portion of space. When V and prePP combine, the portion of the landmark can be, in the case of point adpositions, the location at which the motion takes place (modifier cases), or the location that is the result of the motion (complement cases), or, in the case of path adpositions, the path along which the motion takes place. 7.6 Conclusion This chapter has argued for an analysis of the grammatical constituency of ParCs in terms of a complex predicate with subject and direct object. The analysis is based on (i) the semantic similarities between ParCs and SCVCs, (ii) the semantic similarities between ParCs and constructions from other languages in which motion verbs take a direct object, and (iii) the semantic differences between ParCs and PreCs. The question that now arises is how this analysis relates to the (conflicting) findings of studies that have used constituency tests to determine the grammatical constituency of ParCs. This is the topic of the following chapter.

(ii) [The supersonic aircraft Concorde will make its first test flight from

London on Tuesday since the fleet was grounded following a crash in Paris last year, which killed 113 people.] *British Airways has confirmed that the plane will fly over for nearly three and a half hours before returning to its base.

Chapter 8 Constituency tests revisited 8.1 Introduction My methodology for determining the grammatical constituency of ParCs has been quite different from that of earlier studies. My conclusion that ParCs consist of a separable complex verb, a subject, and a direct object is based on an analysis of the semantics of the construction. Earlier studies, on the other hand, have relied on constituency tests. On the basis of these tests, most studies conclude that the adposition in a ParC shows ambiguous syntactic behavior, which is accounted for in terms of reanalysis: the adposition is essentially a postposition, which is sometimes reanalyzed as a particle.1

The question that this chapter addresses is: do the results from the constituency tests form a problem for the analysis of the grammatical constituency of ParCs presented here? To answer that question, I will first compare the grammaticality judgments reported on in Chapter 2 with actual language usage, because there were quite a few cases of diverging intuitions. Section 8.2 presents attested data of the types that are generally considered to support a ‘particle analysis’, and section 8.3 those that are generally considered to support a ‘postposition analysis’.

Section 8.4 evaluates the significance of these data from a cognitive-grammar perspective. It argues that the crucial question for each set of data should be: does it share the relevant semantics with the ParC,2 i.e. does it designate a motion event in which a trajector traverses a landmark so that result P is achieved? If it does, then the shared semantics can be ascribed to the grammatical constituency that I have argued for on semantic grounds. If it does not, then the set of data is irrelevant for the question of the grammatical constituency of ParCs. Because there are so many constructions involved, I can only suggest an alternative approach here rather than present a full analysis.

1 The conflicting results that have been observed for Dutch ParCs do not constitute an isolated phenomenon. Studies of semantically similar constructions in other languages report similar findings, cf. for example Haig (1981) on Japanese ‘traversal object constructions’ discussed in section 7.3 and Dewell (2000: 55-56) on the direct objects of verbs such as German verlassen ‘leave’ and English leave, depart, and exit. 2 Cf. Langacker’s claim that “symbolic considerations are critical to the determination of grammatical constituency” (1987: 366).

166 Chapter 8

8.2 Attested data supporting a ‘particle analysis’? This section presents authentic examples of constructions that are generally considered to support a ‘particle analysis’. They have been gathered from the 38 million words corpus of the Institute for Dutch Lexicology and from the Internet and are complemented by data that I just happened upon in other sources. The significance of these data for my analysis of the grammatical constituency of ParCs is evaluated in section 8.4.2. What has been considered a real problem for a ‘particle analysis’, first of all, is the idea that ParCs do not have passive counterparts (see Paardekooper 1966, de Haas and Trommelen 1993, Blom 2005). A search of the corpus and the Internet shows, however, that such passive constructions do occur, which ties in with the intuitions found in de Vries (1975) and Hoekstra (1984): (1) Groenendaal werd […] voorbijgereden door […] Jiri Pospisil

(stnov1995) Groenendaal was past-ridden by Jiri Pospisil ‘Groenendaal was overtaken by Jiri Pospisil’ (2) De F1 dacht een reisje naar de dierentuin te hebben gewonnen

the F1 thought a trip to the zoo to have won toen het terrein van Ouwehand werd opgereden3

when the ground of Ouwehand was on/up-driven ‘The F1 thought they had won a trip to the zoo when the grounds of Ouwehand were driven onto’

(3) Tv-Oost had al een soortgelijke reeks, waarbij met een in een auto TV-East had already a similar series, in-which with a in a car

gemonteerde camera de provincie doorgereden werd4 installed camera the province through-driven was ‘TV-East already had a similar series, in which the province was driven through with a camera-equipped car’

(4) Die kwam pas in zicht nadat de hele zee was overgevaren5 that came only in sight after the whole sea was over-sailed ‘That came in sight only after the whole sea had been sailed across’

3 Http://www.hemurenge.nl/verslagen/020906.php, December 2006. 4 Http://www.lindenburg.nl/September2000.htm, December 2007. 5 Http://www.schaakverenigingdetoren.nl, December 2007.

Constituency tests revisited 167

I have not conducted a quantitative study, but passive constructions such as (1)-(4) do not seem to abound. The important point, however, is that they do occur, the significance of which is discussed in section 8.4.2.

A second type of objection to a ‘particle analysis’ is that ParCs do not have nominalized counterparts (Blom 2005). Luif (1992) considers some nominalizations possible, but doubts whether nominalizations with the adpositions in or uit ‘out’ are possible. Just like passive ParCs, such nominalizations do occur, however, which ties in with the intuitions found in Verkuyl and Zwarts (1992) and Beeken (1993): (5) [Het] zal [...] neerkomen op het inrennen van de supermarkt it will down-come to the in-running of the supermarket en het snel omwokken van [...] kant-en-klaar [...] gehakte groenten6 and the quick around-wok of ready-made chopped vegetables

‘It will come down to running into (lit. ‘the running into of’) the supermarket and quickly stirring ready-made chopped vegetables in a wok’

(6) De autobus [...] moet bij het uitrijden van de halteplaats the bus must at the out-driving of the stop-place alle verkeer voorrang verlenen7 all-the traffic priority give ‘The bus must give way to all traffic when exiting (lit. ‘the driving out of’) the station’

(7) Bij het oprijden van de camping by the on/up-driving of the campsite kregen we hier al direct spijt van8

got we here already directly regrets from ‘On entering (lit. ‘the driving onto of’) the campsite we immediately regretted it.’

(8) [...] om tol te mogen heffen voor het overvaren van de rivier9 to toll to be-allowed levy on the over-sailing of the river

‘to be allowed to levy toll on crossing (lit. ‘the sailing over of’) the river’

6 Http://www.trifles.nl/2005/11/08, December 2006. 7 Http://www.peeters.luna.nl/bl19.htm, December 2007. 8 Http://www.camping-frankrijk.nl, December 2006. 9 Http://www.stedplan.nl, December 2006.

168 Chapter 8

(9) pas veel verder, na het doorfietsen van eindeloze lanen met only much further after the through-cycling of endless lanes with

zeer royale vrijstaande huizen komen we weer in het duingebied10 very spacious detached houses come we again in the dune-area ‘Only much further, after having cycled through (lit. ‘the cycling through of’) endless lanes with very spacious detached houses, we find ourselves in the dunes again’

What has also been considered problematic for a ‘particle analysis’

is that ParCs take zijn ‘be’ as their perfect auxiliary rather than hebben ‘have’ (see Paardekooper 1959, de Haas and Trommelen 1993). While it is true that most ParCs take zijn ‘be’ as their perfect auxiliary verb (attested examples are presented in section 8.3), ParCs with hebben do also occur, which ties in with the intuitions found in Hoekstra (1984), Luif (1992), and Blom (2005): (10) Zij heeft als eerste vrouw alle continenten overgelopen11 she has as first woman all continents over-walked ‘She has walked across all continents as the first woman ever’ (11) een groep Spanjaarden die net dat hele kasteel hadden omgelopen12 a group Spaniards who just that entire castle had around-walked ‘a group of Spaniards who had just walked around that entire castle’ (12) nadat we eerst nog een bergje van 25% hadden opgeklommen13 after we first also a mountain-dim of 25% had on/up-climbed ‘but not after we had first climbed a hill of 25% grade’ (13) nadat we de hele buurt hadden doorgereden14 after we the whole neighborhood had through-driven ‘after we had driven through the whole neighborhood’ In addition to their low frequency (again, I have not conducted a quantitative study), some, or even all, of these examples may sound quite marginal to quite a few speakers. When I started to scribble down example (11), from a personal conversation, the person who had said it immediately responded that he should have used waren ‘were’. On my computer screen hadden ‘had’ in (12) is underlined with a squiggly green line; the grammar checker is suggesting that I might be confusing the auxiliaries hebben ‘have’ and zijn

10 Http:// ligfiets.job.coppoolse.com/ligfiets/cyclevision2005, December 2006. 11 Radio 1, Opium, February 23, 2003. 12 From a personal conversation. 13 Http://www.xs4all.nl/~rutty/italy/dag006.htm, December 2007. 14 Http://willemienopcuracao.waarbenjij.nu, December 2007.


‘be’ here. In section 8.4.2 I suggest how the occurrence of such examples can be accounted for on the basis of the semantics of hebben ‘have’ and the semantics of these particular ParCs.

A fourth problem for a ‘particle analysis’ can be found in Blom (2005). She considers topicalization of P+V, i.e. P+V in sentence-initial position, to be impossible. Luif (1992) takes a more moderate position. He believes this type of topicalization to be possible, but only with some adpositions, such as binnen ‘inside’ and voorbij ‘past’, not with in or uit ‘out’. The attested example in (14) suggests, however, that this type of construction does occur, which can be considered support for the ‘particle analysis’: P and V occur in sentence-initial position and therefore form a constituent together, a separable complex verb. (14) Inrijden mag je die weg niet, in-drive may you that road not,

maar dat stukje lopen was geen probleem15 but that bit-DIM walk was no problem

‘You cannot drive into that road, but the short walk was no problem’

Another area in which Luif (1992) has different intuitions depending on the adposition that is used is ‘pronominalization’. What has been considered an argument for a ‘particle analysis’ of ParCs is that unlike the nominal complement of a preposition, the non-subject nominal in a ParC can be a ‘non-r-pronoun’: die ‘that, who’ or dat ‘that’ (see section 2.5 and van Riemsdijk 1978). Such non-r-pronouns are believed to ‘stand in for’ subjects and direct objects (de Schutter 1974, van Riemsdijk 1978, Hoekstra 1984, Luif 1992, de Haas and Trommelen 1993). Here are a number of attested examples: (15) Gillen onder elke tunnel die we doorfietsen16 scream under every tunnel that we through-cycle ‘She screams under every tunnel that we cycle through’

15 Http://www.flitsservice.nl, January 2007. 16 Http://www.verootjuh.nl/?cat=2, December 2006.

170 Chapter 8

(16) De brug die we zojuist overgegaan zijn the bridge that we just over-gone are staat bekend als de Martjesbrug17

stands known as the Martjes-bridge ‘The bridge that we have crossed just now is known as the Martjesbrug’

(17) Maar bij een volgende heuvel die we oplopen but at a next hill that we on/up-walk

zie ik wat er aan de hand is18 see I what there on the hand is ‘But at a next hill that we climb I notice what is going on’

(18) Het was de eerste winkel die we in gingen19 it was the first shop that we in went ‘It was the first shop that we went into’ (19) We [...] komen zo in de Ketelboetstraat die we uitfietsen20 we come so in the Ketelboet-street that we out-cycle

‘That is how we arrive in the Ketelboet Street, which we cycle all the way to the end and out of’

Luif (1992) questions whether non-r-pronouns can occur in constructions with in or uit ‘out’, but see the attested examples in (18) and (19).

Another type of topicalization, but one that has not been called into question, is that of the non-subject nominal. This has also been considered support for the idea that the non-subject nominal and the adposition do not form a constituent: the adposition in such constructions is adjacent to the verb (de Vries 1975, Verkuyl and Zwarts 1992, Haeseryn et al. 1997), cf. attested example (20). (20) drie rivieren moest ie overzwemmen (jgdsep95) three rivers had-to he over-swim ‘three rivers he had to swim across’

For the sake of completeness, I should add the set of attested examples in (21)-(23).

17 Http://www.fietsen.123.nl, December 2006. 18 Http://www.expeditienet.nl/MB07Deloitte, December 2007. 19 Http://onskabouterhuisje.web-log.nl/onskabouterhuisje/2007/01, December 2007. 20 Http://blog.seniorennet.be/snellebirthy, December 2007.


(21) waardoor hij de brug niet over kon rijden (mcnov94ove) because-of-which he the bridge not over could drive ‘because of which he could not drive across the bridge’ (22) terwijl deze [...] het centrum wilde inrijden (mcsep93ove) while this the center wanted in-drive ‘while he wanted to drive into the (town’s) center’ (23) Laten we de eerste bergtop met z’n allen rustig overrijden21 let we the first mountain-top with us all calmly over-ride ‘Let’s calmly ride over the first mountain top together’ They support the widespread intuition that the adposition in a ParC need not follow the non-subject nominal directly (de Schutter 1974, van Riemsdijk 1978, Verkuyl and Zwarts 1992, Beeken 1993, de Haas and Trommelen 1993, Haeseryn et al. 1997, Helmantel 2002, but see Luif 1992: 163, who has his reservations with respect to such constructions with in and uit ‘out’). The adposition of a ParC is, in this respect, similar to the particle of an SCVC, but unlike a preposition, which is directly followed by its nominal complement. A final set of data involves ‘coordination’. Attested data (24)-(26) support Verkuyl and Zwarts’ (1992) and Beeken’s (1993) intuitions.22 They use it as an argument for their ‘particle analysis’, because ordinary separable complex verbs pattern in the same way (but see also the coordination data presented in section 8.3 below).23 (24) Jan gaat de bergen in en ik het oostblok24 Jan goes the mountains in and I the Eastern bloc ‘Jan is going into the mountains and I (into) the Eastern bloc’

21 Studio Sport, Tour de France, 16 July 2000. 22 Note that Verkuyl and Zwarts’ (1992: 394) example in (i) below is a subordinate clause rather than a main clause, which means that in their example the finite verb is clause-final, while in (24)-(26) it is not. (i) omdat Henk het bos inloopt en Frans het park because Henk the wood in-walks and Frans the park ‘because Henk enters the wood and Frans the park’ 23 Cf. for example (i) with the separable complex verb inpakken ‘to wrap’: (i) Thuis pakte Lucy de kado's voor Luchiano in en ik de hare (http:// www.casaspider.com/live/familie_en_vrienden, June 2008) at home packed Lucy the presents for Luchiano in and I (the) hers ‘At home Lucy wrapped the presents for Luchiano and I hers’ 24Http://ambtenarenallerlandenverenigtu.hyves.nl/forum/1030832/SLwu/Wie_is_lid_nr_1200/, June 2008.

172 Chapter 8

(25) jullie reden de oprit op en ik de afrit25 you drove the on-ramp on/up and I the off-ramp ‘you were driving onto the on-ramp and I (onto) the off-ramp’ (26) toen [...] ging gerwin de ene uitgang uit en ik de andere26 then went gerwin the one exit out and I the other-one ‘then Gerwin went out of this exit and I (out of) the other one’

What can be added to these coordination data are attested examples (27)-(29), in which P+V is coordinated with a non-complex verb. (27) De waterweggebruikers moeten the waterway-users must

de zone [...] voorzichtig naderen en voorbijvaren27 the area carefully approach and past-sail ‘The users of the waterway must carefully approach and pass the area’

(28) zodat de bus altijd zonder problemen so that the bus always without problems

de Stadsbrug kan bereiken[...] en oprijden28 the City-bridge can reach and on/up-drive

‘so that the bus can always reach and drive onto the City Bridge without problems’

(29) Bestuurders die een uitrit verlaten of inrijden Drivers who a driveway leave or in-drive

moeten al het overige verkeer laten voorgaan29 must all the other traffic let first-go ‘Drivers who leave or enter a driveway must give way to all other traffic’

This type of coordination is not mentioned in the studies discussed in Chapter 2. If it is assumed that only constituents of the same type can be coordinated, these examples can be considered further support for a ‘particle analysis’.

The conclusion that can be drawn at this point is, in my view, that attested data are crucial for linguistic analysis. Rather than just rely on

25Http://www.bokt.nl/forums/viewtopic.php?t=293464&postdays=0&postorder=asc&start=50, June 2008. 26 Http://roybrinkman89.spaces.live.com/feed.rss, June 2008. 27 Http://ris.vlaanderen.be/ris-beheer/upload/statistiek, December 2007. 28 Http://www2.nijmegen.nl/mmbase/attachments/18046, December 2007. 29 Http://www.goes.nl, December 2007.


intuitions, one should also check what is happening in actual linguistic fact. The question that section 8.4 addresses is what the significance of these data is for my analysis of ParCs in terms of a separable complex verb that takes a subject and a direct object. Can they, from the perspective of Cognitive Grammar, indeed be considered support for such an analysis? We will first, however, focus on attested data of the types that have been considered support for a ‘postposition analysis’. 8.3 Attested data supporting a ‘postposition analysis’? This section presents attested data that correspond to the results of the constituency tests that have been taken as supporting a ‘postposition analysis’. Unlike what we saw in section 8.2, there has been little disagreement about the grammaticality of these kinds of examples, only about those involving topicalization and coordination. It is these types of data that I discuss first. The rest of the section gives an overview of attested data that should not come as a surprise, because there has not been disagreement about those types of data. The aim of the section is to get a good idea of the empirical facts before they are interpreted in section 8.4.3.

After the two types of topicalization discussed in the previous section, taken as support for a ‘particle analysis’, this section discusses a third type. In this type, the non-subject nominal and the adposition occur in topicalized, i.e. sentence-initial, position together, which has been considered support for a ‘postposition analysis’. Paardekooper (1959, 1966), de Vries (1975), Luif (1991), and Haeseryn et al. (1997) consider such topicalizations possible. Their intuitions are supported by the attested example in (30), which describes a detail of a cycling trip. Such an example goes against the intuition of Verkuyl and Zwarts (1992) and Beeken (1993), who mark another example with NP-P in topicalized position with an asterisk and a question mark. (30) De stad in moeten we over een volle driebaans autoweg

the town in must we over a full three-lane highway zonder vluchtstroken rijden, best eng30

without emergency-lanes ride quite scary ‘Into town we have to ride along a busy three-lane highway without emergency lanes, quite scary’

30 Http://www.velomobiel.nl/nl/nieuws_body_0204_nl, December 2006.

174 Chapter 8

The coordination data discussed in the previous section can be complemented by the following. Let me start with Verkuyl and Zwarts’ (1992) intuition that constructions such as (31), from Verkuyl and Zwarts (1992: 394), are ungrammatical, which supports their idea that the adposition and the verb form a separable complex verb. (31) *omdat Henk het bos inloopt en Frans het park in because Henk the wood in-walks and Frans the park in ‘because Henk enters the wood and Frans the park’ Similar examples do occur, however, cf. attested example (32).31 (32) Voor ik het weet loop jij de trap af en ik de trein in32 before I it know walk you the stairs down and I the train in

‘Before I know it, you are walking down the stairs and I into the train’

Moreover, constructions such as (33) and (34), in which two NP-P

sequences are coordinated and which are therefore also quite similar to (31) and (32), seem to occur quite frequently, which ties in with intuitions found in Corver (1990: 301). (33) [...] in paniek rende ik mijn huis uit en de straat op33 in panic ran I my house out and the street on/up ‘In a panic I ran out of my house and onto the street’ (34) Loop de brug over en het Valkenberg park in34 walk the bridge over and the Valkenberg park in ‘Walk across the bridge and into the Valkenberg Park’ For Corver, such examples constitute support for a ‘postposition analysis’, while Beeken argues that this type of coordination is also possible with direct object-particle sequences, cf. her example (35), which was also mentioned in Chapter 2. (35) Hij zou achtereenvolgens moeder op en vader af-bellen.

he would successively mother up and father off-call ‘He would successively call mother up and father off’

31 Notice that the ParC in (32) is a main clause, while (31) is a subordinate clause. 32 Http://watzijzegt.livejournal.com, June 2008. 33 Http://www.m-d-v.nl/pivot/entry.php?id=13, December 2006. 34 Http://lionsconventie2004.nl/homepage.htm, December 2006.


A further set of coordination data that can be added here involves cases in which a prePP is coordinated with an NP-P sequence: (36) De ene keer loop je nog door de rivier

the one time walk you still through the river en dan weer de loopplank over (mcnov92ove) and then again the gangplank over ‘Now you are walking through the river, and then back across the gangplank’

(37) schepen [...] waarmee de conquistadores ships where-with the conquistadores de rivier af en langs de kust naar Mexico voeren35

the river off and along the coast to Mexico sailed ‘ships with which the conquistadores sailed down the river and along the coast to Mexico’

If it is assumed that two coordinated structures must be constituents of the same type, this would be support for a ‘postposition analysis’. As I stated at the beginning of this section, the grammaticality of other data that are considered support for a ‘postposition analysis’ has not been challenged. I will therefore simply list attested examples of such constructions. First of all, there are the construction types in which there is no motion verb with which the adposition could be argued to form a separable complex verb, such as the nominals in (38) and (39): (38) de taxirit de drukke stad uit36 the taxi-ride the busy town out ‘the taxi ride out of the busy town’ (39) het steile pad de berg op37 the steep trail the mountain on/up ‘the steep trail up the mountain’ constructions with met (van Riemsdijk 1978, Luif 1992, Helmantel 2002) such as (40) and (41):

35 Http://nl.wikipedia.org/wiki/Hernando_de_Soto_(ontdekkingsreiziger), December 2007. 36 Peking Express, Net5, September 17, 2006. 37 Http://www.erwinvoogt.com/overland/kkh.html, December 2006.

176 Chapter 8

(40) de vergeetput in met die vent38 the oblivion-well in with that guy ‘into the oubliette with that guy!’ (41) het land uit met die scholiere39 the country out with that student ‘out of the country with that student!’ and constructions in which the NP-P sequence follows a construction that ‘already’ contains a separable complex verb, such as (42) and (43): 40 (42) Prins Bernhard […] ging […] mee de grafkelder in41 prince Bernhard went along the crypt in ‘Prince Bernhard came along into the crypt’ (43) Ze ging me voor de kamer in (Glastra van Loon 1999: 126) she went me before/ahead the room in ‘She showed me the way into the room’ Other constructions in which the adposition cannot be analyzed as forming a complex predicate with the verb are (44)-(47): (44) Heel Palimpsest door duikt hij op (stnov1995) all Palimpsest through dives he on/up ‘Throughout Palimpsest he keeps on popping up’ (45) Een eind de berg op is 1 restaurant42 a distance the mountain on/up is 1 restaurant ´Quite a distance up the mountain is 1 restaurant´ (46) Eenmaal het hek door kom je in een oase van rust43 once the gate through come you in an oasis of calm ‘Once through the gate you arrive in a haven of peace’

38 Http://www.flabber.nl/archief/003292.php, December 2006. 39 Http://www.verblijfopaarde.nl/interactie/oude_columns, December 2006. 40 Other word orders are also possible, cf. e.g.: (i) het zand van het strand ging via de grond het bed mee in

(http://www.vakantiereiswijzer.nl/vakantie/beoordeling/46413/michel, June 2008) the sand from the beach went via the ground the bed with in

‘the sand from the beach got via the floor into our beds with us’ 41 NOS journaal, funeral Prince Claus, October 15, 2002. 42 Http://www.berendbotje.com/bestem/italie/Vaarroute%20Italie, December 2006. 43 Http://www.srtv.info/downloads/Nieuwsbrieven, December 2006.


(47) Eenmaal de brug over gaan we ontbijten44 once the bridge over go we have-breakfast ‘Once over the bridge we go for breakfast’ Since there is no verb in constructions (38)-(47) with which the adposition can be argued to form a separable complex verb, the constructions are considered evidence for a ‘postposition analysis’.

Other constructions whose grammaticality has not been called into question are those involving r-pronouns, caused motion, and the perfect auxiliary zijn ‘be’. (48)-(50) are examples with the r-pronoun waar ‘where, that’. (48) De langste tunnel waar we door rijden the longest tunnel where we through ride

is bijna acht kilometer lang45 is almost eight kilometers long ‘The longest tunnel that we ride through is almost eight kilometers long’

(49) en stort de brug waar we over rijden zowat in46 and throws the bridge where we over ride almost in ‘and the bridge that we ride over almost collapses’ (50) Het laatste kasteel waar we langs liepen was kasteel Terworm47 the last castle where we along walked was castle Terworm ‘The last castle that we walked by was Castle Terworm’ Such examples have been used to argue for a ‘postposition analysis’ (van Riemsdijk 1978, Hoekstra 1984, de Haas and Trommelen 1993). Others have argued, however, that such data do not correspond to what I call ParCs, but to PreCs (de Schutter 1974, Luif 1992, Helmantel 2002). The grammaticality of constructions with r-pronouns has, in other words, not been called into question, but their significance for the issue at hand has been debated, a matter that will be taken up in section 8.4.3 below.

(51) and (52) are attested examples of caused motion constructions discussed in section 2.9.

44 Http://spinazie.blogspot.com/2006_09_01_spinazie_archive, December 2006. 45 Http://home.wanadoo.nl/hilgers/noordkaap, December 2006. 46 Http://www.omzwerving.nl/topic/2/page/7, December 2006. 47 Http://www.henrifloor.nl/w20010521.htm, December 2007.

178 Chapter 8

(51) Ik reed mijn motor [...] de boot op48 I rode my motorbike the boat on/up ‘I rode my motorbike onto the boat’ (52) Ze duwde hem de kamer uit49 she pushed him the room out ‘She pushed him out of the room’ They have been considered to show that the nominal preceding the adposition in these constructions and in ParCs cannot be a direct object (Luif 1992, de Haas and Trommelen 1993, Blom 2005).50 The argument goes as follows: because mijn motor ‘my motorbike’ is the direct object in (51), de boot ‘the boat’ cannot be a direct object. And finally, as has already been noted in section 8.2, ParCs take the perfect auxiliary zijn ‘be’ more often than hebben ‘have’. Attested examples are (16), repeated here, (53) and (54): (16) De brug die we zojuist overgegaan zijn the bridge that we just over-gone are staat bekend als de Martjesbrug51

stands known as the Martjes-bridge ‘The bridge that we have crossed just now is known as the Martjesbrug’

(53) dat de advocaat met opzet het kanaal is ingereden52 that the lawyer on purpose the canal is in-driven ‘that the lawyer drove into the canal on purpose’ (54) Ik ben vol vertrouwen het ijs opgestapt (mcdec95spo)

I am full-of confidence the ice on/up-stepped ‘I stepped onto the ice full of confidence’

This has been taken as support for the view that such constructions are different from constructions with transitive SCVs, which usually take hebben.

48 Http://www.motorclubschoonebeek.nl/reisverslag.htm, August 2008. 49 Http://www.literairethrillers.nl/?id=463, December 2006. 50 Hoekstra (1984: 172), on the other hand, suggests that the structure of the “NP P sequence […] may be ambiguous”. 51 Http://www.fietsen.123.nl/route%20hoogland%20west%20natuur.htm, December 2006. 52 NRC, 11 July 2001.


The attested data presented in this section are of the types that have been used to support a ‘postposition analysis’. Section 8.4.3 addresses the question of whether such data form a problem for my analysis of the grammatical constituency of ParCs in terms of a separable complex verb which takes a subject and a direct object. 8.4 The significance of the data from the constituency tests 8.4.1 A cognitive-grammar perspective The question that this section addresses is: what is the significance of the attested data presented in sections 8.2 and 8.3 for the analysis presented in Chapter 7? This question can be reformulated in terms of two subquestions. The first is: should the data presented in section 8.2 be considered support for a particle analysis, i.e. for the analysis that I have presented? This question is addressed in subsection 8.4.2. The second subquestion is: should the data presented in section 8.3 be considered support for a postposition analysis, i.e. are they problematic for the analysis that I have presented? This question is addressed in subsection 8.4.3. The remainder of the present subsection briefly discusses the status of the attested examples presented in 8.2 and 8.3 and lays out the aspects of Cognitive Grammar that are relevant for answering these questions. Sections 8.2 and 8.3 have shown that types of constructions that have been considered to be ‘impossible’ or ‘ungrammatical’ by some do in fact occur. What should we make of this? Note, first of all, that the constituency tests discussed in Chapter 2 are based on sentences in isolation. Fauconnier (1994: xlv-xlvi) points out the risks of considering sentences in isolation when doing semantics:

When a sentence is examined in isolation, and its interpretations are studied, it is necessary to implicitly construct a discourse in which to interpret it. By default, a minimum discourse is usually chosen, with the implication that this will yield the ‘real’, ‘core’, context-independent meaning of the sentence. This implication is unwarranted; there is no reason why the particular configuration associated with a linguistic expression in a minimum discourse should contain the defining characteristics for the meaning potential of that expression in any discourse.

The methodology also poses a risk when trying to establish whether a construction is possible or not, cf. Fauconnier (1994: xxvii):

180 Chapter 8

native speakers’ judgments in isolation do not inform us about meaning and form per se, but rather about the subject’s ability to construct appropriate minimum contexts.

I consider it a real possibility that the speakers who reject, for example, passive ParCs in isolation will find attested examples of such constructions quite acceptable when considered in their natural contexts (cf. also Langacker 1987: 36-37). There may, however, be attested examples that are acceptable and natural to certain speakers, yet sound odd to others, even when the examples are considered in their natural contexts. Some of the ParCs with hebben ‘have’ presented in 8.2 seem to be good examples of this. This is quite understandable from the perspective of Cognitive Grammar, which defines ‘the grammar of a language’ (Langacker 1987: 57) as

those aspects of cognitive organization in which resides a speaker’s grasp of established convention. […] It can be characterized as a structured inventory of conventional linguistic units.

Examples such as those with hebben ‘have’ apparently have a limited ‘degree of conventionality’ (cf. Langacker’s objection to the term ‘grammaticality’, 1987: 66). What we are concerned with here, however, is their meaning.53 The attested examples presented in sections 8.2 and 8.3 show, in other words, that speakers produce these types of constructions, which they apparently consider to be meaningful, whatever their degree of conventionality. As was described in sections 5.4.1 and 7.5.2, Cognitive Grammar does not distinguish a separate level of syntactic structure (Langacker 1987: 53-54):

the only structures permitted in the grammar of a language […] are (1) phonological, semantic, or symbolic structures that actually occur in linguistic expressions; (2) schemas for such structures; and (3) categorizing relationships involving the elements in (1) and (2).

Purely syntactic constituency, i.e. without phonological or semantic content, is therefore not recognized in the theory: “no descriptive constructs are

53 Cf. Langacker (2000: 54) on constructions such as ?Under the bed is all dusty and ?Near the fire is warmer:

Though marginal for some, expressions like these do occur and are readily accepted by many speakers. It is in any case not their degree of conventionality that is relevant here, but rather their meaning.


permitted that lack both phonological and semantic content” (Langacker 1987: 54). This also means that syntactic derivations, on which the constituency tests in the previous section are largely based, are not allowed in the theory: “overt structures cannot be derived from hypothetical ‘underlying’ structures having a substantially different character” (Langacker 1987: 54).

A passive construction such as (1), for example, is not derived somehow from a structure such as active (55). (1) Groenendaal werd […] voorbijgereden door […] Jiri Pospisil

(stnov1995) Groenendaal was past-ridden by Jiri Pospisil ‘Groenendaal was overtaken by Jiri Pospisil’ (55) Jiri Pospisil reed Groenendaal voorbij Jiri Pospisil rode Groenendaal voorbij ‘Jiri Pospisil overtook Groenendaal’ Nor is a nominal such as the one in bold in (7), for example, analyzed as a syntactic derivation from an underlying structure such as (56). (7) Bij het oprijden van de camping by the on/up-driving of the campsite kregen we hier al direct spijt van54


(56) Toen de politie die camping opreed (mcoct95ove) when the police that campsite on/up-drove ‘When the police drove onto that campsite’ Instead, they are alternate constructions, i.e. complex symbolic structures, with their own semantics, cf. Langacker (1987: 39):

Grammar (like lexicon) embodies conventional imagery. By this I mean that it structures a scene in a particular way for purposes of linguistic expression [...]. Two roughly synonymous sentences with the same content words but different grammatical structures – including, in particular, sentences generally analyzed as being transformationally related – are claimed instead to be semantically distinct by virtue of their different grammatical organization per se.

54 Http://www.camping-frankrijk.nl, December 2006

182 Chapter 8

We may wonder whether from such a theoretical perspective any construction can be used as evidence at all for a certain constituency analysis of another construction. It is possible, but it requires a careful analysis of the semantics of the constructions involved. Remember from Chapter 7 that grammatical constituency is defined in Cognitive Grammar as “the order in which simpler symbolic structures are successively integrated to form progressively more elaborate ones” (Langacker 2000: 149). Therefore, only constructions that share the relevant semantics with the type of ParCs that I have discussed can be used as evidence for a certain constituency analysis. The ‘relevant semantics’ that I am referring to here is the constraint that ParCs impose on the interpretation process: a ParC designates a motion event in which a trajector traverses a landmark so that result P is achieved: the trajector moves from not P to completely P. If that semantic structure is shared, both constructions can be analyzed as consisting of a complex predicate. The subsections below argue that passive and nominalized constructions such as (1) and (7) are such constructions, as are constructions with NP-P in ‘topicalized’ position. Constructions with r-pronouns, however, are not. 8.4.2 The significance of the data presented in 8.2 My evaluation of the significance of the data presented in section 8.2 will be limited to a discussion of constructions with hebben (as opposed to zijn), nominalizations, and passives. I suggest that the other types of data discussed in that section can be dealt with along the same lines, i.e. data with either the non-subject nominal or P+V in topicalized position, data with non-r-pronouns, and data in which the non-subject nominal is not adjacent to the adposition.

From the fact that ParCs usually take zijn ‘be’ as their auxiliary and occasionally hebben ‘have’ I do not conclude that adpositions in constructions with zijn are postpositions, while those in constructions with hebben are particles. Such a conclusion is drawn in analyses in terms of structural ambiguity/reanalysis (cf. e.g. Luif 1992: 165, Blom 2005: 268-272). My analysis of these data is different. Regardless of whether zijn or hebben is used, a ParC designates a motion event in which a trajector traverses a landmark so that result P is achieved. The use of hebben or zijn has, therefore, no implications for the analysis presented in Chapter 7 that ParCs consist of a complex separable verb that takes a subject and a direct object, which is based on this analysis of their semantics.

As I stated in Chapter 5, I analyze hebben ‘have’ and zijn ‘be’ as imposing different construals on a motion event. With hebben the emphasis


is on the type of act, while with zijn the focus is on the change of location (cf. Geerts et al. 1984, Honselaar 1987, Shannon 1990, and Haeseryn et al. 1997). Since ParCs designate motion events in which a trajector changes location, zijn is a natural choice. If, however, such a motion event involving a trajector changing location is construed as a type of act, this can be designated by means of hebben.

This can account nicely for the attested examples of ParCs with hebben in (10)-(13), repeated here. As I said, what they share semantically with ParCs with zijn is that the trajector has traversed a landmark so that result P is achieved. In (10)-(13), however, each of the motion events constitutes a formidable achievement, note the use of modifiers such as alle ‘all’, hele ‘entire’, and the daunting gradient percentage in (12). The traversal is, in other words, construed as a type of act, which explains the use of hebben ‘have’ here. (10) Zij heeft als eerste vrouw alle continenten overgelopen 55 she has as first woman all continents over-walked ‘She has walked across all continents as the first woman ever’ (11) een groep Spanjaarden die net dat hele kasteel hadden omgelopen56 a group Spaniards who just that entire castle had around-walked ‘a group of Spaniards who had just walked around that entire castle’ (12) nadat we eerst nog een bergje van 25% hadden opgeklommen57 after we first also a mountain-DIM of 25% had on/up-climbed ‘but not after we had first climbed a hill of 25% grade’ (13) nadat we de hele buurt hadden doorgereden58 after we the whole neighborhood had through-driven ‘after we had driven through the whole neighborhood’

The use of zijn or hebben is, in other words, not diagnostic for the grammatical constituency of the constructions: they all consist of a separable complex verb with a subject and a direct object. Instead, the auxiliaries impose their own construals on the motion event that is designated by the construction. One reason why ParCs with hebben have a low frequency is that in most cases, a change of location is just that, a change of location, and not a change of location construed as a type of act. A further reason for the low frequency could be that this is a case of language change, or at least some kind of unstable situation, in which only some speakers use the 55 Radio 1, Opium, February 23, 2003. 56 From a personal conversation. 57 Http://www.xs4all.nl/~rutty/italy/dag006.htm, December 2007. 58 Http://willemienopcuracao.waarbenjij.nu, December 2007.

184 Chapter 8

construction with hebben when change of location is construed as a type of act, while others use zijn.

The attested examples of ‘nominalized’ ParCs and passive ParCs are support for the analysis presented in Chapter 7. I do not want to say that because a ParC such as (56) can be nominalized, see attested example (7), it behaves syntactically like a construction with a separable complex verb, see (57) and (58). (56) Toen de politie die camping opreed (mcoct95ove) when the police that campsite on/up-drove ‘When the police drove onto the campsite’ (7) Bij het oprijden van de camping by the on/up-driving of the campsite kregen we hier al direct spijt van59


(57) dat ik mijn huiswerk afmaakte60 that I my homework off-made ‘that I finished my homework’ (58) Griffel hechtte grote waarde aan het afmaken van huiswerk61 Griffel attached great value to the off-making of homework ‘Griffel felt very strongly about finishing homework’ Instead, the finite clauses in (56) and (57) and the constructions in bold in (7) and (58) are different constructions with their own distinct semantics. What we can say, however, is that the construction in bold in (7) shows semantic similarities to (56): oprijden (lit. ‘on/up-driving’) in (7) evokes a conceptualization of the same type as opreed (‘drove onto’) in (56); this conceptualization is the result of the conceptual integration of the semantics of the adposition and the semantics of the verb of motion.62 We can observe the same for (57) and (58): afmaakte ‘finished’ and afmaken ‘finishing’ evoke a conceptualization that is the result of integrating the semantics of the particle af ‘off’ and the semantics of the verb maken ‘make’. In this sense, the ParCs in (56) and (7) and the SCVCs in (57) and (58) are similar

59 Http://www.camping-frankrijk.nl/, December 2006. 60 Http://www.svj.hvu.nl/toestand/020530.pdf, December 2006. 61 Http://www.dreuzels.com/forum/forum_posts.asp?TID=15096, December 2006. 62 A difference is that oprijden in (7) has a nominal profile (it does not profile time), while opreed in (56) has a temporal relational profile, i.e. it is ‘anchored in time (cf. Langacker 1991a: 33).


semantically, which can be accounted for, as I have, in terms of similar grammatical constituency.

In the same way, I do not want to say that because the active ParC in (55) can be passivized, see attested example (1), it behaves syntactically like an SCVC, cf. (57) and (59). (55) Jiri Pospisil reed Groenendaal voorbij Jiri Pospisil rode Groenendaal voorbij ‘Jiri Pospisil overtook Groenendaal’ (1) Groenendaal werd […] voorbijgereden door […] Jiri Pospisil

(stnov1995) Groenendaal was past-ridden by Jiri Pospisil ‘Groenendaal was overtaken by Jiri Pospisil’ (57) dat ik mijn huiswerk afmaakte63 that I my homework off-made ‘that I finished my homework’ (59) Hij vraagt of dat huiswerk is afgemaakt64 he asks whether that homework is off-made ‘He is asking whether that homework has been finished’ Instead, active and passive constructions are two different types of constructions with their own semantics. At the same time, however, the constructions in (55) and (1) share similar semantics: the trajector Jiri Pospisil traverses (part of the ‘perimeter’ of the) landmark Groenendaal (another cyclist) so that the result is achieved that Pospisil is completely voorbij ‘past’ Groenendaal. This semantic structure can be accounted for in terms of a complex predicate (reed … voorbij in (55) and voorbijgereden in (1)): verb and adposition integrate semantically so that voorbij ‘past’ is understood as the result of a motion event. That passive ParCs are, presumably, less frequent than passive SCVCs should not be taken to indicate that active ParCs, or the majority of active ParCs, do not consist of a complex predicate that takes a subject and direct object. The alternative explanation that Cognitive Grammar offers is that the semantic structure of the ParC is such that only in special cases is it compatible with the semantics of the Passive Construction. In (1), for example, the landmark of the ParC is a cyclist who is affected by the motion event because he is overtaken by the trajector, i.e. he has less of a chance to win the race. If an active construction has an affected direct object, it is

63 Http://www.svj.hvu.nl/toestand/020530.pdf, December 2006. 64 Http://www.nationaalpark.nl/docs/200603131337295821.doc, June 2008.

186 Chapter 8

likely to have an acceptable passive ParC counterpart. The direct object of a ParC, however, is usually not affected at all: traversal usually does not effect a change.65 While hebben and zijn cannot be considered arguments for or against a ‘particle analysis’, the occurrence of passive and nominalized ParCs can be so considered. They share with active ParCs the type of semantics that formed the basis for proposing a ‘particle analysis’. My impression is that the data in 8.2 that have not been discussed here are similar to passive and nominalized ParCs. That is, data with topicalized P+V or topicalized NP, non-r-pronouns, or non-adjacency of NP and P have their own semantics, yet share with ParCs such (55) or (56) that they impose the ParC constraint described in Chapter 6, on the basis of which they too are analyzed as complex predicates with a subject and a direct object. 8.4.3 The significance of the data presented in 8.3 This subsection suggests that the data presented in 8.3, which are generally taken as supporting a ‘postposition analysis’, are in fact not problematic for the present analysis of ParCs in terms of an SCV with a subject and a direct object. Three types of data are dealt with in some detail: data with r-pronouns, data with NP-P in topicalized positions, and ‘caused-motion data’.

The non-subject nominal of a ParC is considered by some to be replaceable by an r-pronoun, which is considered support for the idea that nominals are complements of a postposition (van Riemsdijk 1978, Hoekstra 1984, de Haas and Trommelen 1993), cf. attested examples (48)-(50): (48) De langste tunnel waar we door rijden the longest tunnel where we through ride

is bijna acht kilometer lang66 is almost eight kilometers long ‘The longest tunnel that we ride through is almost eight kilometers long’

65 Affectedness is not the whole story as can be seen from examples (2)-(4), in which the traversed entities (designated by the subjects) do not seem to be affected by the traversal. Another feature of the passive construction is that the participant that carries out the process designated by the verb can be left unspecified (cf. Vanden Bosch 1992 and Cornelis 1997 for analyses of Dutch passives; see also Rice 1987 for English passives). 66 Http://home.wanadoo.nl/hilgers/noordkaap, December 2006.


(49) en stort de brug waar we over rijden zowat in67 and throws the bridge where we over ride almost in ‘and the bridge that we ride over almost collapses’ (50) Het laatste kasteel waar we langs liepen was kasteel Terworm68 the last castle where we along walked was castle Terworm ‘The last castle that we walked along was Castle Terworm’ Others have argued, however, that data such as (48)-(50) do not correspond to what I call ParCs, but to PreCs (de Schutter 1974, Luif 1992, Helmantel 2002): (59) We rijden door de tunnel ‘We are riding through the tunnel’ (60) We rijden over de brug ‘We are riding across the bridge’ (61) We liepen langs het kasteel ‘We walked along the castle’ The constructions in (48)-(50) are, in other words, irrelevant for the question of the grammatical constituency of ParCs, because they do not share the relevant semantics.

I would like to add something to this discussion of constructions with r-pronouns. My suggestion is that some of these constructions with r-pronouns do involve a separable complex predicate. In such cases the r-pronoun does not, however, replace the non-subject nominal, i.e. the direct object. Instead, it designates a more global ‘setting’ (Langacker 2000: 67) where the motion event designated by the complex verb takes place. A clear example of this is (62): (62) Uit het slaapkamerraam kijkt Lisa [...] naar de rivier, out-of the bedroom-window looks Lisa to the river waar een fuut zo stil als een papieren bootje voorbijglijdt,

where a grebe as quietly as a paper boat-DIM past-glides naar het dampende grasland daarachter (Enquist 1994: 77)

towards the steaming grasslands behind it ‘From the bedroom window Lisa looks out at the river, where a grebe, as quietly as a paper boat, glides by, towards the steaming grasslands behind it’

67 Http://www.omzwerving.nl/topic/2/page/7, December 2006. 68 Http://www.henrifloor.nl/w20010521.htm, December 2007.

188 Chapter 8

Waar ‘where’ in (62) does not elaborate the landmark of voorbij ‘past’, but designates a region that functions as a more global setting, i.e. de rivier ‘the river’. In this setting, a motion event takes place that can be described in its simplest form by means of the following construction: (63) een fuut glijdt voorbij ‘A grebe glides by’ This motion event is designated by a complex predicate, voorbijglijden ‘glide by’, which requires two participants, a trajector and a landmark. The trajector is elaborated by the subject, but there is no linguistic element that elaborates the landmark. We have already seen such examples in Chapter 7; they are repeated here: (64) zodat de pont kan overvaren (wk199318) so that the ferry can over-sail ‘so that the ferry can sail across’ (65) Zodra de chauffeur is ingestapt (jgdoct93) as-soon-as the driver is in-stepped ‘As soon as the driver has gotten in’ In (64) and (65) the landmark can be inferred from the context and need not be expressed. In the case of (63) the landmark is a mental reference point,69 i.e. the point (or region) in or near the river that Lisa is looking at. The grebe traverses that point (or region) so that the result is that the grebe is completely past it.

Constructions with r-pronouns cannot, in other words, be used to argue for a postposition analysis, because the r-pronoun does not correspond semantically to the non-subject nominal of a ParC. Others have already argued that it corresponds to the complement of a preposition. I have added another possibility, in which the r-pronoun designates a setting. The latter type of interpretation is, for example, possible for (49) above.

Another type of construction that has been used to argue for a postposition analysis of ParCs is illustrated by (51) and (52):

69 See for the notion ‘mental reference point’ Langacker’s (2000) analysis of across discussed in section 4.3.6.


(51) Ik reed mijn motor [...] de boot op70 I rode my motorbike the boat on/up ‘I rode my motorbike onto the boat’ (52) Ze duwde hem de kamer uit71 she pushed him the room out ‘She pushed him out of the room’ As has been stated in section 8.3, such constructions have been taken as an argument against the direct-object status of the non-subject nominal in a ParC (Luif 1992, de Haas and Trommelen 1993, Blom 2005, but see Hoekstra 1984: 172): since, for example, mijn motor ‘my motorbike’ is the direct object in (51), de boot ‘the boat’ cannot be a direct object too.

Notice, however, that the semantics of such caused-motion constructions are to a large extent similar to that of ParCs. The role of the referent of the nominal preceding the adposition (i.e. de boot ‘the boat’ in (51) and de kamer ‘the room’ in (52)) is identical to the role of the referent of the non-subject nominal in a ParC: it is (completely or partially) traversed by another entity. The role of the referent of the nominal following the verb (i.e. mijn motor ‘my motorbike’ in (51) and hem ‘him’ in (52)) is similar to that of the subject referent of a ParC: it traverses the landmark from where it is not P to where it is completely P. The difference is that in the caused-motion constructions this participant’s motion is necessarily caused motion. The ‘traversing entity’ of a caused-motion construction is, in other words, a patient: it is affected by the process.

I consider both non-subject nominals in these caused-motion constructions, i.e. the ‘traversing entity’ and the ‘traversed entity’, to be direct objects, be it different types of direct objects. Analyzing the ‘traversing entity’ as a direct object is uncontroversial: it is a prototypical direct object in the sense that it is a patient (cf. Langacker 1991a: 321-323), which makes it a possible subject of a passive construction. Calling the ‘traversed entity’ a direct object is controversial. What is clear, however, is that its role is the same as that of the non-subject nominals of ParCs, which Chapter 8 argued to be direct objects. They are not, of course, direct objects of the prototypical kind, but they do fit the broader definition of direct objects provided by Cognitive Grammar: “a noun phrase whose profile corresponds to the landmark of the process profiled by the clause as a whole” (Langacker 1995: 154). Following the literature on the Japanese constructions discussed in 7.3, we may want to call such nominals ‘traversal

70 Http://www.motorclubschoonebeek.nl/reisverslag.htm, August 2008. 71 Http://www.literairethrillers.nl/?id=463, December 2006.

190 Chapter 8

objects’ (Kuno 1973, Martin 1975, Haig 1981, Beavers 2006) to distinguish them from ‘prototypical direct objects’.

Crucial to direct-objecthood (of the prototypical or traversal object kind) is that the process designated by the verb is conceptually dependent on the participant that the direct object designates. Section 7.5.3 has proposed for ParCs that integration of the verb and the adposition results in the conceptualization of a motion event that has P as its result and that depends conceptually on two participants. For constructions such as (50) and (51), I suggest that the verb and the adposition, too, integrate semantically in such a way that P is understood as the result of the event.72 Because the verbs involved in such constructions have caused-motion semantics (or, in the case of rijden ‘drive’, are compatible with caused-motion interpretations), this event does not require two, but three participants, i.e. it depends conceptually on three participants, which are expressed by three bare nominals: a subject and two direct objects, i.e. one of the prototypical type and one traversal object.

The third and final type of data to be discussed in this section can be illustrated by (30), which describes a detail of a cycling trip: (30) De stad in moeten we over een volle driebaans autoweg

the town in must we over a full three-lane highway zonder vluchtstroken rijden, best eng73

without emergency-lanes ride quite scary ‘Into town we have to ride along a busy three-lane highway without emergency lanes, quite scary’

Haeseryn et al. (1997: 508-509) provide a similar example, i.e. (66), as an argument for analyzing in as a postposition in (67): de garage in ‘into the garage’ occurs in topicalized position and forms, therefore, a constituent. (66) De garage in rijdt hij altijd zelf the garage in drives he always himself ‘Into the garage he always drives himself’ (67) Hij rijdt de garage in he drives the garage in ‘He is driving into the garage’

72 The details of the integration are a matter of future research. 73 Http://www.velomobiel.nl/nl/nieuws_body_0204_nl, December 2006.


While I agree that de stad in ‘into town’ in (30) and de garage in ‘into the garage’ in (66) are constituents, I do not consider in a postposition. Instead, I analyze such NP-P sequences as ParC ‘fragments’: they evoke the same semantics as ‘full’ ParCs, but profile only parts of it. The similar semantics can be explained if it is assumed that ParC fragments and full ParCs are similar compositionally, i.e. P is interpreted as the result of a process (which in the case of (30) is unprofiled because unexpressed). This resultative process requires two participants, a trajector and a landmark. The trajector in (30) is not expressed, but understood; the landmark is expressed. I consider (30) and (66) to be semantically similar, to a large extent, to an example such as (44). (44) Heel Palimpsest door duikt hij op (stnov1995) all Palimpsest through dives he on/up ‘Throughout Palimpsest he keeps on popping up’ The clause-initial NP-P sequence evokes in such cases a similar motion event as is designated by a full ParC, be it more schematically. Heel Palimpsest door ‘throughout Palimpsest’ in (44) evokes as its base a (subjective) motion event (cf. the discussion of subjective motion in section 4.3.6) in which the trajector (the conceptualizer) ‘traverses’ Palimpsest, a book (or rather, the conceptualization of the story), from beginning to end. Even though there is no verb of motion present, a schematic process of motion is evoked with which the semantics of door ‘through’ conceptually integrates. The compositional path of the fragment heel Palimpsest door is, in other words, the same as that of a full ParC, which by cognitive-grammar definition means that they share the same grammatical constituency. De stad in ‘into town’ in (30) similarly evokes as its base a schematic motion event, but here it is a case of objective motion. Just like heel Palimpsest door, the fragment evokes, in my view, the whole motion event ‘on its own’,74 i.e. it does so without the verb rijden in the remainder of the example. The unspecified trajector partially traverses de stad ‘the town’ from one of the town’s boundaries to where it is completely in the town, i.e. completely contained in the town. It is a schematic motion event because the trajector, the manner of motion, and time are left unprofiled. On the basis of previous context, however, we can fill these details in.

74 Cf. Verhagen’s (1996: 793-794) claim that “whenever two elements in a sentence are distinguished as separate, the one that comes first is to be conceptualized independently with respect to the one that follows (whatever aspect of meaning is at order)”; see also Pardoen (1998: 434).

192 Chapter 8

The processes that are designated by the rest of sentences (30) and (44), i.e. the sentences without the NP-P sequences, take place while the motion events evoked by the NP-P sequences are unfolding. While the conceptualizer in (44) mentally progresses through Palimpsest, a certain character keeps popping up. Similarly in (30), while the trajector, or rather trajectors (a group of cyclists), move into town, they are involved in the type of process described by the rest of the sentence: moeten we over een volle driebaans autoweg zonder vluchtstroken rijden ‘we have to ride over a busy three-lane highway without emergency lanes’. What is crucial is that the ‘ParC semantics’ in (30), i.e. traversal from not in to completely in is due to the fragment de stad in ‘into town’. The remainder of the sentence does not have the ‘ParC semantics’. Instead, it is a PreC: rijden ‘ride’ is ‘purely’ an intransitive verb; it does not require a direct object. Notice in this respect the scope of moeten ‘have to’: (30) is not about having to ride into town, but about having to ride over a busy three-lane highway (while going into town). The other examples of NP-P in sentence-initial position presented in section 8.3 are (45)-(47): (45) Een eind de berg op is 1 restaurant75 a distance the mountain on/up is 1 restaurant ´Quite a distance up the mountain is 1 restaurant´ (46) Eenmaal het hek door kom je in een oase van rust76 once the gate through come you in an oasis of calm ‘Once through the gate you arrive in a haven of peace’ (47) Eenmaal de brug over gaan we ontbijten77 once the bridge over go we have-breakfast ‘Once over the bridge we go for breakfast’ They too evoke schematic motion events and are therefore analyzed in the same way, i.e. as ParC fragments. They do show a difference to the ParCs in (30) and (44). In (30) and (44) the process designated by the remainder of the sentence takes place as the motion event designated by the NP-P sequence unfolds. In (45)-(47), on the other hand, the process designated by the remainder of the sentence takes place when the motion event designated by the NP-P has been completed, i.e. when result P has been achieved. Notice that the subjective/objective motion distinction can be observed in

75 Http://www.berendbotje.com/bestem/italie/Vaarroute%20Italie, December 2006. 76 Http://www.srtv.info/downloads/Nieuwsbrieven, December 2006. 77 Http://spinazie.blogspot.com/2006_09_01_spinazie_archive, December 2006.


these examples too. (45) is a case of subjective motion: the conceptualizer mentally scans the path from the foot of de berg ‘the mountain’ to somewhere op ‘on’ the mountain.78 (46) and (47) are cases of objective motion.

We have seen that Verkuyl and Zwarts (1992) present their intuition that NP-P topicalization is not possible or questionable as support for their ‘particle analysis’, while others consider NP-P topicalization possible and use that as support for a ‘postposition analysis’. My perspective differs from both these positions. I have found attested examples of ‘topicalized’ NP-P sequences, but I do not conclude that this is evidence for a postposition analysis. Instead, I have observed that in such cases the semantic structure of the NP-P sequence is similar to that of a ParC, only more schematic. It evokes as its base a motion event with P as its result, in which a trajector traverses a landmark from not P to completely P. The difference with a full ParC is that in a ParC fragment only the result and the landmark are profiled. Still, P is interpreted as the result of a schematic process of motion and therefore a particle.

I expect that the alternative approach that I have sketched here for constructions with r-pronouns, cases of NP-P in topicalized position and for caused-motion constructions such as (50) and (51) opens up a new way of accounting also for the data from 8.3 that were not discussed here, such as coordination and NP-P sequences in nominal constructions such as (38) and (39), in met-constructions, and in constructions which ‘already’ contain a SCV. 8.5 Conclusion This chapter has presented attested examples of the types of constructions discussed in earlier studies that have addressed the question of the grammatical constituency of ParC. As it turns out, all types of constructions occur in actual language usage, also those that were considered ‘ungrammatical’ or ‘impossible’ in some studies. This finding underscores, in my view, the importance of using authentic language material.

78 Cf. de Schutter’s (1974: 284, my translation from the Dutch original, MB) observation about an example similar to (45) that

the construction with a postposition, including the one in [(45)], has by itself, as an essential part of its constructional semantics, a directional aspect; there is not the least bit of doubt about the reality of the observation that noticing the [restaurant] in [(45) …] is implicitly related to a spectator who moves de berg op ‘up the mountain’.

194 Chapter 8

Six types of data were then considered in some more detail. The fact that ParCs can take zijn, and in special cases hebben, was argued to be irrelevant for the grammatical constituency question, as was the occurrence of constructions with r-pronouns. The auxiliaries impose their own construal on the motion event expressed by the complex predicate. The r-pronoun is not semantically similar to the non-subject nominal ParC. Instead, it corresponds to the complement of a preposition, or it designates a setting for a ParC in which the landmark is left unspecified. The occurrence of passive and nominalized ParCs was argued to be support for the analysis presented in Chapter 7, while the occurrence of NP-P in sentence-initial position and certain caused-motion constructions were argued not to pose a problem for this analysis. These constructions share with ‘ordinary’ ParCs the semantics that formed the basis for proposing the analysis in Chapter 7.

Chapter 9 Conclusion 9.1 A summary of the analysis This study has answered a long-standing question in Dutch linguistics: what is the grammatical constituency of constructions such as (1)? (1) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) a girl cycles the Doctor Larij-road on/up ‘A girl is cycling onto the Doctor Larij Road’ They consist of a complex predicate (P+V), a subject, and a certain type of direct object (a traversal object). This is why they have been referred to as ParCs in this study.

This conclusion is based on a methodology that is quite different from that of previous attempts to determine the grammatical constituency of ParCs. They rely on such constituency tests as passivization, topicalization, and auxiliary choice, which, as Chapter 2 argues, leave the question unresolved. The present study has argued for an analysis of the grammatical constituency of ParCs on the basis of their semantics. The theoretical background for this has been the cognitive-grammar view of grammatical constituency as “the order in which simpler symbolic structures are successively integrated to form progressively more elaborate ones” (Langacker 2000: 149).

To come to a semantic analysis of ParCs, I have contrasted ParCs with PreCs, which can consist of the same words as ParCs, but in which the adposition precedes the non-subject nominal, as in e.g. (2). (2) Een jongen [… fietste] op de Rijksparallelweg (mcnov93ove) a boy cycled on the national-parallel-road ‘A boy was cycling on the National Parallel Road’ Helmantel (2002), the most elaborate analysis of the semantic differences between ParCs and PreCs to date, was shown in Chapter 3 to leave quite a number of questions unanswered. Chapter 6 proposes that the semantics of ParCs can be characterized as follows: a ParC profiles a motion event in

196 Chapter 9

which a trajector traverses1 a landmark so that result P is achieved: the trajector moves from where it is not P to where it is completely P. While PreCs also profile motion events, they need not be resultative, nor do they require that the trajector traverses the landmark or ends up being completely P.

Chapter 7 argues that these semantic differences between ParCs and PreCs can be explained if it is assumed that ParCs and PreCs are compositionally different. ParCs were argued to be compositionally more similar to resultative separable complex verb constructions (SCVCs) such as (3) than to PreCs. (3) de gevangene haalt de trekker over (jnlfeb93) the prisoner pulls the trigger over ‘The prisoner pulls the trigger’ ParCs and SCVCs are similar in that (i) P and V are semantically integrated so that they designate an event of which P is understood as the result, (ii) this event necessarily requires two participants, which are expressed as subject and direct object, (iii) the event progresses from not P to completely P, and (iv) that the event takes place ‘where the referent of the direct object is’, i.e. it progresses through the direct object. Furthermore, ParCs are semantically similar to constructions in other languages in which motion verbs take a direct object. A number of such constructions have been analysed in terms of Dowty’s (1991) notion of Incremental Path Theme, a subtype of his Incremental Theme. Section 7.3 argued that the notion Incremental Path Theme is problematic. In such constructions with motion verbs and direct objects (e.g. cross the desert), there is not just one participant which can be taken into account to conclude anything about the progression of the event. Instead, both participants, the referents of the subject and the direct object, need to be taken into account. What the direct objects of such constructions do share with other direct objects, such as those in mow the lawn or eat an apple, is that the event progresses through the direct object. A third semantic parallel that was drawn is that between the semantic differences between ParCs and PreCs and the holistic/partitive effect observed in argument/oblique alternations. The effect can be observed for ParCs and PreCs too: not in the direct object vs. the complement of the preposition, but in the subjects. In a ParC, the subject is completely P as a

1 Cf. section 6.2 for how ‘traversal’ is defined in this study: it includes complete, partial, and perimeter traversal.

Conclusion 197

result of the motion event, while in a PreC, the subject need just be partially P.

On the basis of these semantic parallels, section 7.5 has argued that ParCs are compositionally different from PreCs. The adposition of a ParC forms a complex predicate with the verb. Together they designate a motion event of which P is understood as the result. The event requires two participants, a trajector and a landmark, which are expressed by the subject and the direct object in an active clause.

The adposition of a PreC, on the other hand, forms a prepositional-phrase construction with the nominal that follows it directly in the phonological structure. The nominal’s profile elaborates the schematic landmark in the semantic structure of the adposition. This has a distinctive semantic effect that is responsible for the ‘partitive’ interpretation that has often been observed for nominal complements of prepositions. The elaboration of the landmark has the effect that the prePP includes the specification that a portion of that landmark is involved with the trajector: a point in the case of point adpositions (e.g. in ‘in’ and op ‘on’) and a path in the case of path adpositions (e.g. door ‘through’ and over ‘over, across’). Chapter 5 examined the ways in which such prePPs integrate semantically with a verb of motion. The chapter concludes that two types of PreCs need to be distinguished, but that the auxiliaries hebben and zijn cannot be used as diagnostic for the two types (pace e.g. Hoekstra 1984, Hoekstra and Mulder 1990, van Hout 1996, and Helmantel 2002): they cross-cut the distinction. In the case of point adpositions, there are two possibilities. Either the prePP’s trajector is elaborated by the process of motion designated by the verb, or the prePP specifies a point that elaborates the endpoint of the path in the semantics of the verb of motion. For most PreCs with path adpositions, however, there is no need to distinguish two types, which is a second departure from earlier analyses: the prePP specifies a path that elaborates the path in the semantics of the verb of motion. The idea of an endpoint arises in interpretation. 9.2 General conclusions A basic tenet of cognitive linguistics is that semantic structure is conceptualization, in which construal, i.e. the way in which a conceptualization is structured, plays a crucial role. The importance of this is supported by the present analysis. Different constructions can be used to describe the same event in the world; still, they are different in meaning because they portray that event in different ways. The study has given several examples of this. A PreC (or ParC) with the auxiliary hebben, for

198 Chapter 9

example, focuses on the motion as a type of activity, while such a construction with zijn focuses on the motion as change of location. A ParC with in or op profiles motion that (partially) traverses the landmark, while a Complement PreC with in or op profiles motion that ends in or op the landmark without requiring or focusing on traversal of the landmark. A third example is that in ParCs with in or op, the trajector’s motion with respect to the landmark can continue as long as it is construed as leading to the result that the trajector is completely in or on the landmark. Speakers’ expectations play an important role in this. A second basic tenet of cognitive linguistics is that language expressions underdetermine the rich interpretations that are associated with them, i.e. the idea of ‘partial’ rather than ‘strict’ compositionality. We draw on contextual knowledge in the widest sense of the word to come up with an interpretation for an expression, i.e. to integrate the meanings of the words. The present study has explored an, in my view, more specific version of this tenet, namely the idea of interpretation as constraint satisfaction (Fauconnier 1990, 1994, 1997, Verhagen 1997, 2000). Interpreting a language expression is a highly creative process that is constrained by the (constructional and lexical) semantics of the expression and its context. These constructional, lexical, and contextual constraints work in parallel.

The study has explored the idea that the semantic structure of a word or a grammatical construction is a single constraint, i.e. a single, abstract conceptualization that ‘maintain[s ... its] identity in different contexts’ (Verhagen 1997: 7). In this it deviates from the common cognitive-linguistic assumption that polysemy is the norm in both lexical and constructional semantics. I consider the approach to semantics in terms of a (set of) constraint(s) to be fruitful. It has resulted in an analysis of Dutch over in terms of a single constraint, presented in Chapter 4, as well as an analysis of the ParC in terms of a single constraint. Chapter 4 argues that what has been overlooked in previous analyses of Dutch over (or English over) is its force-dynamic character. The constraint proposed for Dutch over does not specify whether the trajector is higher than the landmark or not: that is a matter of context. Instead it specifies that there is a force pointing from the path to the landmark. What is also a matter of context is how the trajector is related to the landmark by means of the mental path: it can move along the path, it can be an elongated, static object that extends along the path, or it can be located at the end of the path. The study underscores the importance of basing linguistic analyses on attested examples. Quite a number of grammaticality judgments that the constituency tests rely on turn out not to be in accordance with empirical fact. With regard to a construction’s range of interpretations, too, it turns out

Conclusion 199

that more is possible than one may think on the basis of constructed examples and introspection. From the perspective of interpretation as constraint satisfaction, this is quite understandable. An interpretation of a ParC is based on lexical, constructional, and contextual constraints. With constructed examples, one runs the risk of only thinking of default contexts and overlooking generalizations (cf. Fauconnier 1994: xxvii, xlv-xlvi).

A final conclusion is that the cognitive-grammar view of grammatical constituency has opened up a new way to analyze ParCs. The present study has combined the cognitive-grammar notion of grammatical constituency as compositional path with the idea that the semantic structure of a word is a single constraint that maintains its identity across contexts, including constructional contexts. The semantic differences between ParCs and PreCs were therefore hypothesized to be ascribable to the different hierarchical ways of integrating the semantics of the words involved.

This has led to a more unified and insightful analysis of ParCs. While earlier analyses conclude, for example, that in ParCs with hebben, the adposition is a particle, but in ParCs with zijn, the adposition is a postposition, the present analysis concludes that the adposition is a particle in either construction. The adposition is understood as the result of the motion event that is designated by the adposition and the verb together. Hebben and zijn do not correlate with different constituent structures, but ‘simply’ impose different construals on the motion event. The grammatical constituency of a construction can only be analyzed if the semantics of the construction is taken into account. 9.3 Suggestions for further research At different points in this study I have identified issues that require further research, such as non-spatial uses of adpositions (section 4.1, fn. 1), the relation between extraposition and auxiliary choice (section 5.4.2, fn. 18), ParCs with uit ‘out’ and af ‘off, down’ with ‘length modifiers’ (section 6.2.3), ParCs with langs ‘along’, voorbij ‘past’, om ‘around’, and rond ‘around’ (section 6.2.4), and the semantics of the different constructions presented in Chapter 8. This final section brings up some other issues for which the findings and general approach of the present study can serve as a theoretical and descriptive starting point.

ParCs with verbs that do not designate motion. I have considered only ParCs with verbs of motion, but other verbs occur in ParC-like constructions too: the verb zijn ‘be’, modals, certain verbs of visual perception, and certain verbs that designate activities that can be carried out while moving, see (3)-(6).

200 Chapter 9

(3) toen die ook het lokaal uit was (Peper 409) when that-one also the room out was ‘when she had also left the room (was also out of the room)’ (4) Iedereen die de stad in wil [...] (NRC, April 4, 2005) everyone who the town in wants ‘Everyone who wants to go into town’ (5) Toen keek je zo die knie binnen2 then looked you so that knee inside ‘Then you could look right into that knee’ (6) [...] zag ik hem nog net in de verte de straat uitsnuffelen3 saw I him still yet in the distance the street out-sniff ‘I could just see him sniffing (his way) out of the street’

The development of the ParC. What has often been observed is (i) that Middle Dutch did not have what I call ParCs (Hogenhout-Mulder 1983: 74, van der Sijs 2005: 114, Cloutier 2006: 70), and (ii) that what is expressed by ‘postpositions’ in Modern Dutch is expressed by prepositions in Middle Dutch (Gerritsen 1978: 37, Hogenhout-Mulder 1983: 74, van der Wal and van Bree 1992: 150, Cloutier 2006: 70), see for instance Middle Dutch (7) and Modern Dutch (8) from Gerritsen (1978: 37, boldface mine, MB): (7) en ghinck met hastechheiden op den berghe in de stat ‘and went hastily up the mountain to the city’ (8) en ging met haast de berg op naar de stad ‘and went hastily up the mountain to the city’ Blom (2005: 269, 273), however, does discuss a small number of constructions with what she calls postpositions from Middle Dutch, such as (9) and (10). (9) Ende hi voer alle dlant dore4 and he drove all the-land through ‘And he drove through the whole country’ (10) Sijn predicaren sullen alle die werelt over wanderen5 his preachers shall all the world over wander ‘His preachers will wander all over the world’

2 René Eikelkamp in Sportpaleis De Jong, September 24, 1999. 3 From a personal conversation. 4 Bijbelvertaling 1360 (1460-1462, South Western Limburg). 5 Tafel vanden kersten ghelove – Zomerstuk (1480, Holland).

Conclusion 201

It would be interesting to see how the ParC has developed. What might be taken into account in such a study is constructions such as (11)-(13) from Middle Dutch, which no longer occur in Modern Dutch, but were still around in the nineteenth century.6 (11) Men sleiptene vord ende weder Te berge op, te dale neder7 people dragged-him to and fro at mountain up at valley down ‘People dragged him to and fro, up the mountain, down the valley’ (12) Dat ysere quam ten rugge dore8 that-the iron came at-the back through ‘that the iron came through the back’ (13) Met dese sprake ginck Uracla metten Grave ter poorten uyt9 with these words went Uracla with-the Count at-the gate out ‘With these words Uracla went out of the gate with the Count’ In these constructions the adposition follows a noun that is preceded by te, ten, or ter, which is the preposition te constracted with a case-marked definite article. Has the development of the ParC in some way been reinforced by the existence of such constructions? It may be the case that the contraction of te with a definite article is at some point no longer recognized as such, but understood as simply de, i.e. just the definite article.

Cross-linguistic comparisons. The study presented here fits in a large body of (cognitive-linguistic) research into the expression and conceptualization of spatial relations. As such it can serve as a basis for cross-linguistic comparisons in the tradition of Talmy (1983), Levinson (1992), and Bowerman (1996a, 1996b). The difference between ParCs and PreCs with in and op ‘on’, for example, may at first sight seem quite similar to that between the German prepositions in and auf ‘on’ taking dative vs. accusative case, cf. (14) and (15) from Smith (1987: 120).

6 Cf. e.g. (i) from Marcellus Emants’ Op reis naar Zweden (1877, retrieved from the Digital Library for the Dutch Letters, http://www.dbnl.nl). (i) [Hij] stapte met kalme waardigheid ter deure uit he stepped with calm dignity at-the door out ‘He stepped out of the door with calm dignity’ De poort uit ‘out of the gate’, de deur uit ‘out of the door’, het raam uit ‘out of the window’, which are analyzed as metonymous in Chapter 6 (see fn. 17), are possibly ‘remnants’ of this construction. 7 From Lanceloet en het hert met de witte voet (1300, retrieved from the Digital Library for the Dutch Letters, http://www.dbnl.nl). 8 Grimbergse oorlog, 1300. 9 De historie van Partinoples, grave van Bleys, 1250.

202 Chapter 9

(14) Mein Freund fährt in der Stadt ‘My friend drives (around) in the[DAT] city’ (15) Er fährt in die Stadt ‘He drives into the[ACC] city’ Closer inspection reveals, however, that while Dutch particles correspond in such examples with accusative case in German, Dutch prepositions cover an area that is larger than that of dative case in German (cf. also Draye 1992), cf. (16)-(18) from Smith (1987: 106-107). (16) Das Buch liegt auf dem Tisch ‘The book is lying on the[DAT] table’ (17) Hans [legt] das Buch auf den Tisch ‘Hans [lays] the book on(to) the[ACC] table’ (18) Er begleitete mich auf die Strasse

‘He accompanied me onto the[ACC] street (he led me out onto the street)’

The corresponding constructions from Dutch are provided in (19)-(21). A prePP is used in (20), because what is at issue is that at the end of the (caused) motion the book is on the table; traversal of the table is not required here. In (21), on the other hand, there is traversal from a boundary of the street to where the trajector is on the street, and the adposition follows de straat ‘the street’. (19) Het boek ligt op de tafel. (20) Hans legt het boek op de tafel. (21) Hij begeleidde me de straat op. Whether there is traversal of the landmark or not, an issue that is critical for the use of Dutch ParCs vs. PreCs, does not appear to be important in the case of the German expressions.10

Circumpositions. Finally, it is for me an entirely open question as to how to analyze constructions such as (22)-(26), which contain so-called ‘circumpositions’ (see Haeseryn et al. 1997: 526-530 for an overview of

10 Cf. Smith’s (1987: 107) characterization of the difference between the dative and the accusative in these cases:

The DAT versions […] describe situations in which the TR of the preposition continually satisfies the configurational requirements of the preposition, while the ACC versions describe situations where these requirements are not initially fulfilled.

Conclusion 203

such combinations). Compared with ParCs and PreCs, these constructions contain one more adposition: there is an adposition preceding the non-subject nominal and one following it. (22) Zes raketten vlogen recht op hun doel af (jgdsep92) six rockets flew straight on their target off ‘Six rockets were flying straight at their target’ (23) Hij zwom tegen de stroom in (wb94-4) he swam against the current in ‘He was swimming against the current’ (24) De bestuurder [...] liep voor zijn auto langs (mcnov92spo) the driver walked in-front-of his car along ‘The driver was passing the front of his car’ (25) als je net onder de brug door bent gevaren (wk199422) when you just under the bridge through are sailed ‘when you have just passed under the bridge’ (26) Twee getuigen zwommen naar de man toe (wk199503) two witnesses swam to the man to ‘Two witnesses swam to the man’ I suggest that what is required to determine the grammatical constituency of such expressions is a semantic analysis of the type proposed in this study for ParCs and PreCs, on the basis of which an analysis of their grammatical constituency can be proposed.

References Beavers, John, 2006. Argument/oblique alternations and the structure of

lexical meaning. Ph.D. dissertation, Stanford University. Beeken, Jeannine, 1993. Spiegelstructuur en variabiliteit: Pre- en

postposities in het Nederlands. Leuven: Peeters. Beliën, Maaike, 2002. Force dynamics in static prepositions: Dutch aan, op,

and tegen. In: Hubert Cuyckens and Guenther Radden, eds., Perspectives on prepositions, 195-209. Tübingen: Niemeyer.

Beliën, Maaike, 2006. Uit: Meer plaats dan pad. Voortgang Jaarboek voor de Neerlandistiek 24: 25-35.

Bennett, David C., 1975. Spatial and temporal uses of English prepositions: An essay in stratificational semantics. London: Longman.

Blom, Corrien, 2005. Complex predicates in Dutch: Synchrony and diachrony. Ph.D. dissertation, Vrije Universiteit Amsterdam.

Booij, Geert, 1998. Samenkoppelingen en grammaticalisatie. In: E. Hoekstra and C. Smits, eds., Morfologiedagen 1996, 6-20. Amsterdam: Meertens Instituut.

Bowerman, Melissa, 1996a. The origin of children’s spatial semantic categories: Cognitive versus linguistic determinants. In: John J. Gumperz and Stephen C. Levinson, eds., Rethinking linguistic relativity, 145-176. Cambridge: Cambridge University Press.

Bowerman, Melissa, 1996b. Learning how to structure space for language: A crosslinguistic perspective. In: Paul Bloom, Mary A. Peterson, Lynn Nadel, and Merrill F. Garrett, Language and space, 385-436. Cambridge, MA: MIT Press.

Brugman, Claudia Marlea, 1983. The story of over. Bloomington, IN: Indiana University Linguistics Club.

Burzio, Luigi, 1984. Italian syntax: A government-binding approach. Dordrecht: Reidel.

Capelle, Bert, 2004. The particularity of particles, or why they are not just ‘intransitive prepositions’. In Hubert Cuyckens, Walter de Mulder, and Tanja Mortelmans, eds., Adpositions of movement, 29-57. Amsterdam: Benjamins.

Choi-Jonin, Injoo and Laure Sarda, forthcoming. Transitive motion verbs in French and Korean. In: H. Cuyckens, W. De Mulder, T. Mortelmans, and M. Goyens., eds., Variation and change in adpositions of movement. Amsterdam: Benjamins.

206

Cloutier, Robert Allen, 2006. Adpositional phrases in the history of Dutch: The case of in. ACLC Working Papers 1: 67-77.

Cornelis, Louise H., 1997. Passives and perspective. Amsterdam: Rodopi. Corver, Norbert, 1990. The syntax of left branch extractions. Ph.D.

dissertation, Katholieke Universiteit Brabant. Croft, William and D. Alan Cruse, 2004. Cognitive linguistics. Cambridge:

Cambridge University Press. Cuyckens, Hubert, 1991. The semantics of spatial prepositions in Dutch: A

cognitive-linguistic exercise. Ph.D. dissertation, University of Antwerp.

Cuyckens, Hubert, 1994. Family resemblance in the Dutch spatial preposition op. In: Monika Schwarz, ed., Kognitive Semantik/Congitive Semantics: Ergebnisse, Probleme, Perspektiven, 179-195. Tübingen: Narr.

de Haas, Wim and Mieke Trommelen, 1993. Morfologisch handboek van het Nederlands: Een overzicht van de woordvorming. The Hague: SDU.

den Hertog, C.H., 1973. Nederlandse spraakkunst. Amsterdam: Versluys. de Schutter, G., 1974. De Nederlandse zin: Poging tot beschrijving van zijn

structuur. Brugge: De Tempel. de Vries, Jan W., 1975. Lexicale morfologie van het werkwoord in modern

Nederlands. Leiden: Universitaire Pers Leiden. Dewell, Robert B., 1994. Over again: Image-schema transformations in

semantic analysis. Cognitive Linguistics 5(4): 351-380. Dewell, Robert B., 2000. Case meaning and sequence of attention: Source

landmarks as accusative and dative objects of the verb. In: Ad Foolen and Frederike van der Leek, eds., Constructions in Cognitive Linguistics, 47-65. Amsterdam: Benjamins.

Dewell, Robert B., 2007. Moving around: The role of the conceptualizer in semantic interpretation. Cognitive Linguistics 18(3): 383-415.

Dowty, David R., 1991. Thematic proto-roles and argument selection. Language 67(3): 547-619.

Draye, Luc, 1992. Zum Trajektiv: Ein Kapitel aus einer kognitiv orientierten niederländisch-deutschen Kontrastivgrammatik. Leuvense Bijdragen 81, 163-203.

Fauconnier, Gilles, 1990. Invisible meaning. BLS 16: 390-404. Fauconnier, Gilles, 1994. Mental spaces: Aspects of meaning construction in

natural language. Cambridge: Cambridge University Press. Fauconnier, Gilles, 1997. Mappings in thought and language. Cambridge:

Cambridge University Press.

207

Fillmore, Charles J., 1982. Frame semantics. In: The Linguistic Society of Korea, ed., Linguistics in the morning calm, 111-137. Seoul: Hanshin.

Fillmore, Charles J., 1985. Frames and the semantic of understanding. Quaderni di Semantica 4(2): 222-254.

Freyd, Jennifer J., Teresa M. Pantzer and Jeannette L. Cheng, 1988. Representing statics as forces in equilibrium. Journal of Experimental Psychology: General 117(4): 395-407.

Geeraerts, Dirk, 1992. The semantic structure of Dutch over. Leuvense Bijdragen 81: 205-230.

Geerts, G., W. Haeseryn, J. de Rooij, and M.C. van den Toorn, 1984. Algemene Nederlandse spraakkunst. Groningen: Wolters-Noordhoff.

Gehrke, Berit, 2007. On directional readings of locative prepositions. Proceedings of ConSOLE XIV: 99-120.

Gerritsen, Marinel, 1978. De opkomst van SOV patronen in het Nederlands inverband met woordvolgordeveranderingen in de Germaanse talen. In: J.G. Kooij, ed., Aspekten van woordvolgorde in het Nederlands, 3-39. Leiden: Vakgroep Nederlandse Taal- & Letterkunde.

Haeseryn, W., K. Romijn, G. Geerts, J. de Rooij, and M.C. van den Toorn, 1997. Algemene Nederlandse spraakkunst. Groningen: Nijhoff.

Haig, John H., 1981. Are traversal objects objects? Papers in Linguistics: International Journal of Human Communication 14(1): 69-101.

Hagoort, Peter and Jos van Berkum, 2007. Beyond the sentence given. Philosophical transactions of the Royal Society B: Biological Sciences 362 (nr. 1481): 801-811.

Helmantel, Marjon, 1998. Simplex adpositions and vector theory. The Linguistic Review 15: 361-388.

Helmantel, Marjon, 2002. Interactions in the Dutch adpositional domain. Ph.D. dissertation, Leiden University.

Hoekstra, Teun, 1984. Transitivity: Grammatical relations in government- binding theory. Dordrecht: Foris.

Hoekstra, Teun and René Mulder, 1990. Unergatives as copular verbs: Locational and existential predication. The Linguistic Review 7: 1-79.

Hogenhout-Mulder, Maaike, 1983. Cursus Middelnederlands. Groningen: Wolters-Noordhoff.

Honselaar, Wim, 1987. Zijn vs. hebben in het samengesteld perfectum. De Nieuwe Taalgids 80(1): 55-68.

Hopper, Paul and Sandra Thompson, 1980. Transitivity in grammar and discourse. Language 56(2): 251-299.

208

Jackendoff, Ray, 1983. Semantics and cognition. Cambridge, MA: MIT Press.

Janssen, Theo A.J.M., 2003. Monosemy versus polysemy. In: Hubert Cuyckens, René Dirven, and John R. Taylor, eds., Cognitive approaches to lexical semantics, 93-122. Berlin: Mouton de Gruyter.

Janssen, Theo A.J.M., 2007. A speaker/hearer-based grammar: The case of possessives and compounds. In: Mike Hannay and Gerard Steen, eds., Structural-functional studies in English grammar, 353-387. Amsterdam: Benjamins.

Johnson, Mark, 1987. The body in the mind. Chicago, IL: The University of Chicago Press.

Koopman, Hilda, 1997. The structure of Dutch PPs. Ms., UCLA. Kraak, Albert and Willem G. Klooster, 1968. Syntaxis. Culemborg: Stam-

Kemperman. Kreitzer, Anatol, 1997. Multiple levels of schematization: A study in the

conceptualization of space. Cognitive Linguistics 8(4): 291-325. Kuno, Susumo, 1973. The structure of the Japanese language. Cambridge,

MA: MIT Press. Lakoff, George, 1987. Women, fire, and dangerous things: What categories

reveal about the mind. Chicago, IL: The University of Chicago Press.

Langacker, Ronald W., 1987. Foundations of Cognitive Grammar, vol. I: Theoretical prerequisites. Stanford, CA: Stanford University Press.

Langacker, Ronald W., 1991a. Foundations of Cognitive Grammar, vol. II: Descriptive application. Stanford, CA: Stanford University Press.

Langacker, Ronald W., 1991b. Concept, image, and symbol: The cognitive basis of grammar. Berlin: Mouton de Gruyter.

Langacker, Ronald W., 1995. Conceptual grouping and constituency in Cognitive Grammar. In: Ik-Hwan Leek, ed., Linguistics in the Morning Calm 3, 149-172. Seoul: Hanshin.

Langacker, Ronald W., 1997. Constituency, dependency, and conceptual grouping. Cognitive Linguistics 8(1): 1-32.

Langacker, Ronald W., 2000. Grammar and conceptualization. Berlin: Mouton de Gruyter.

Levin, Beth, 1993. English verb classes and alternations: A preliminary investigation. Chicago, IL: The University of Chicago Press.

Levin, Beth and Malka Rappaport Hovav, 1995. Unaccusativity: At the syntax-lexical semantics interface. Cambridge: MIT Press.

Levinson, Stephen C., 1992. Language and cognition: The cognitive consequences of spatial description in Guugu Yimithirr . Working

209

Paper 13, Cognitive Anthropology Research Group, Max Planck Institute for Psycholinguistics, Nijmegen.

Lieber, Rochelle and Harald Baayen, 1997. A semantic principle of auxiliary selection. Natural Language and Linguistic Theory 15: 789-845.

Lindner, Susan Jean, 1983. A lexico-semantic analysis of English verb particle constructions with out and up. Bloomington, IN: Indiana University Linguistics Club.

Luif, Jan, 1992. Over richtingsbepalingen. In: E.C. Schermer-Vermeer, W.G. Klooster, and A.F. Florijn, eds., De kunst van de grammatica, 157-167. Amsterdam: Vakgroep Nederlandse Taalkunde, Universiteit van Amsterdam.

Luif, Jan, 2000. De betekenis van adposities in plaats- en richtingsbepalingen: Een verkenning. In: Hans den Besten, Els Elffers, and Jan Luif, eds., Samengevoegde woorden, 181-192. Leerstoelgroep Nederlandse Taalkunde, Universiteit van Amsterdam.

Mackenzie, J. Lachlan, 1992. English spatial prepositions in Functional Grammar. Working Papers in Functional Grammar 46.

Martin, Samuel, 1975. A reference grammar of Japanese. New Haven, CT: Yale University Press.

Muehleisen, Victoria and Mutsumi Imai, 1997. Transitivity and the incorporation of ground information in Japanese path verbs. In: Marjolijn Verspoor, Kee Dong Lee, and Eve Sweetser, eds., Lexical and syntactical constructions and the construction of meaning, 329-346. Amsterdam: Benjamins.

Paardekooper, P.C., 1959. Voor- en achterzetsels. De Nieuwe Taalgids 52: 310-320.

Paardekooper, P.C., 1966. Beknopte ABN-syntaksis. Den Bosch: Malmberg. Pardoen, Justine A., 1998. Interpretatiestructuur: Een onderzoek naar de

relatie tussen woordvolgorde en zinsbetekenis in het Nederlands. Ph.D. dissertation, Vrije Universiteit Amsterdam.

Perlmutter, D.M., 1978. Impersonal passives and the Unaccusative Hypothesis. BLS 4: 157-189.

Regier, Terry, 1996. The human semantic potential. Cambridge, MA: MIT Press.

Rice, Sally, 1987. Towards a cognitive model of transitivity. Ph.D. dissertation, University of California, San Diego.

Rice, Sally, Dominiek Sandra, and Mia Vanrespaille, 1999. Prepositional semantics and the fragile link between space and time. In: Masako K. Hiraga, Chris Sinha, and Sherman Wilcox, eds., Cultural,

210

psychological and typological issues in Cognitive Linguistics, 108-127. Amsterdam: Benjamins.

Sandra, Dominiek and Sally Rice, 1995. Network analyses of prepositional meaning: Mirroring whose mind – the linguist’s or the language user’s? Cognitive Linguistics 6(1): 89-130.

Schermer-Vermeer, E.C., 1970. Syntaxis op transformationeel-generatieve grondslag. Levende talen: 85-101.

Shannon, Thomas F., 1990. The unaccusative hypothesis and auxiliary choice. In: Henning Andersen and Konrad Koerner, eds., Historical Linguistics 1987, 461-488. Amsterdam: Benjamins.

Smith, Michael B., 1987. The semantics of dative and accusative in German: An investigation in Cognitive Grammar. Ph.D. dissertation, University of California, San Diego.

Sorace, Antonella, 2000. Gradients in auxiliary selection with intransitive verbs. Language 76(4): 859-890.

Talmy, Leonard, 1983. How language structures space. In: H. Pick and L. Acredolo, eds., Spatial orientation: Theory, research, and application, 225-282. New York: Plenum Press.

Talmy, Leonard, 1985. Lexicalization patterns: Semantic structure in lexical forms. In: Timothy Shopen, ed., Language typology and syntactic description, vol. III: Grammatical categories and the lexicon, 57-149. Cambridge: Cambridge University Press.

Talmy, Leonard, 1988. Force dynamics in language and cognition. Cognitive Science 12: 49-100.

Talmy, Leonard, 2000. Towards a cognitive semantics, vol. I: Concept structuring systems. Cambridge, MA: MIT Press.

Tenny, Carol L., 1992. The aspectual interface hypothesis. In: Ivan A. Sag and Anna Szablocsi, eds., Lexical matters, 1-27. Stanford, CA: CSLI Publications.

Tenny, Carol L., 1994. Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer.

Tomasello, Michael, 1999. The cultural origins of human cognition. Cambridge, MA: Harvard University Press.

Tyler, Andrea and Vyvyan Evans, 2001. Reconsidering prepositional polysemy networks: The case of over. Language 77(4): 724–765.

Tyler, Andrea and Vyvyan Evans, 2003. The semantics of English prepositions. Cambridge: Cambridge University Press.

Vandeloise, Claude, 1991. Representation, prototypes, and centrality. In: S.L. Tsohatzidis, ed., Meanings and prototypes: Studies in linguistic categorization, 403-437. London: Routledge.

211

Vandenbosch, Luc, 1992. Aspekten van passiefvorming in het Nederlands: Een kognitief-pragmatische benadering. Ph.D. disseration, University of Antwerp.

Vanden Wyngaerd, Guido, 2001. Measuring events. Language 77(1): 61-90. Van der Gught, Fieke, Klaas Willems, and Ludovic De Cuypere, 2007. The

iconicity of embodied meaning: Polysemy of spatial prepositions in the cognitive framework. Language Sciences 29: 733-754.

van der Leek, Frederike, 1996. The English conative construction: A compositional account. Chicago Linguistics Society 32: 363-378.

van der Leek, Frederike, 2000. Caused motion and the ‘bottom-up’ role of grammar. In: Ad Foolen and Frederike van der Leek, eds., Constructions in Cognitive Linguistics, 301-331. Amsterdam: Benjamins.

van der Leek, Frederike and Ad Foolen, 2006. De conceptuele basis van doordat, omdat en want. Voortgang Jaarboek voor de Neerlandistiek 24: 73-84.

van der Sijs, Nicoline, 2005. De geschiedenis van het Nederlands in een notendop. Amsterdam: Bert Bakker.

van der Wal, Marijke and Cor van Bree, 1992. Geschiedenis van het Nederlands. Utrecht: Het Spectrum.

van Hout, Angeliek, 1996. Event semantics of verb frame alternations: A case study of Dutch and its acquisition. Ph.D. dissertation, Katholieke Universiteit Brabant.

van Riemsdijk, Henk C., 1978. A case study in syntactic markedness: The binding nature of prepositional phrases. Lisse: De Ridder.

van Staden, Miriam, Melissa Bowerman, and Mariet Verhelst, 2006. Some properties of spatial description in Dutch. In: Stephen C. Levinson and David P. Wilkins, eds., Grammars of space: Explorations in cognitive diversity, 475-511. Cambridge: Cambridge University Press.

Verhagen, Arie, 1996. Sequential conceptualization and linear structure. In Eugene H. Casad, ed., Cognitive linguistics in the redwoods: The expansion of a new paradigm in linguistics, 793-817. Berlin: Mouton.

Verhagen, Arie, 1997. Context, meaning, and interpretation, in a practical approach to linguistics. In: Leo Lentz and Henk Pander Maat, eds., Discourse analysis and evaluation: Functional approaches, 7-39. Amsterdam: Rodopi.

Verhagen, Arie, 2000. Achter het Nederlands: Over interacties tussen taal en achtergrondcognitie. Inaugural lecture, March 24, 2000, Leiden University.

212

Verkuyl, Henk, 1993. A theory of aspectuality: The interaction between temporal and atemporal structure. Cambridge: Cambridge University Press.

Verkuyl, Henk and Frans Zwarts, 1992. Postpositie, incorporatie en terminativiteit. In: E.C. Schermer-Vermeer, W.G. Klooster, and A.F. Florijn, eds., De kunst van de grammatica, 383-400. Amsterdam: Vakgroep Nederlandse Taalkunde, Universiteit van Amsterdam.

Zaenen, Annie, 1993. Unaccusativity in Dutch: Integrating syntax and lexical semantics. In: James Pustejovsky, ed., Semantics and the lexicon, 129-161. Dordrecht: Kluwer.

Zwarts, Joost, 2006. Om en rond: Een semantische vergelijking. Nederlandse Taalkunde 11(2): 101-123.

Zwarts, Joost, forthcoming. Forceful prepositions. In: Vyvyan Evans and Paul Chilton, eds., Language, cognition, and space: The state of the art and new directions. London: Equinox.

Sources de Winter, Leon, 1989. Kaplan. Amsterdam: De Bezige Bij. Enquist, Anna, 1994. Het meesterwerk. Amsterdam: De Arbeiderspers. Glastra van Loon, Karel, 1999. De passievrucht. Amsterdam: L.J. Veen. Peper, Rascha, 2003. Wie scheep gaat. Amsterdam: L.J. Veen. Schmidt, Annie M.G., 1971. Pluk van de Petteflet. Amsterdam: Querido.

Samenvatting in het Nederlands Constructies, beperkingen en voorstellingswijze: Adposities in het Nederlands De term adpositie wordt gebruikt voor partikels, preposities (voorzetsels) en postposities (achterzetsels). De adpositie op in zin (1), bijvoorbeeld, is een partikel: op vormt samen met dronk een scheidbaar samengesteld werkwoord (opdrinken). (1) Iedereen dronk zijn drankje op (mcmar95ove) (2) Een jongen [… fietste] op de Rijksparallelweg (mcnov93ove) In (2) is op een prepositie: het gaat vooraf aan de nominale woordgroep de Rijksparallelweg waarmee het een constituent (zinsdeel) vormt. Een dergelijke constituent, op de Rijksparallelweg, wordt een prepositional phrase (prePP) genoemd. De centrale vraag in dit proefschrift is een kwestie die de Nederlandse taalkunde al decennia lang bezighoudt: wat is de constituentstructuur van zinnen als (3)? (3) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) Sommige taalkundigen hebben voorgesteld dat op in dit type zinnen een postpositie is: het vormt een constituent met de nominale woordgroep die eraan voorafgaat. Een dergelijke constituent, de Dokter Larijweg op, wordt een postpositional phrase (postPP) genoemd. Anderen analyseren op als een partikel: het vormt samen met fietst een scheidbaar samengesteld werkwoord (opfietsen), waarvan de Dokter Larijweg het direct object (lijdend voorwerp) is. Weer anderen analyseren dit type zinnen als syntactisch ambigu: volgens hen is op in essentie een postpositie, maar kan die soms geheranalyseerd worden als een partikel.

Dat eerdere studies naar de constituentstructuur van zinnen als (3) tot verschillende conclusies komen, ligt niet aan een verschil in methodologie. Om de constituentstructuur te bepalen maken ze gebruik van syntactische tests zoals passivisatie, topicalisatie en hulpwerkwoordskeuze. Deze tests zijn echter niet zonder problemen, zoals Hoofdstuk 2 laat zien. Sommige taalkundigen denken bijvoorbeeld dat zinnen als (3) niet passief gemaakt kunnen worden (*De Dokter Larijweg wordt opgefietst). Ze

216 Samenvatting in het Nederlands

concluderen op basis daarvan dat de Dokter Larijweg geen direct object kan zijn van een scheidbaar samengesteld werkwoord opfietsen, maar dat op een postpositie is en de Dokter Larijweg op een postPP. Anderen delen het idee dat zinnen als (3) niet passief gemaakt kunnen worden, maar beschouwen dat niet als een argument tegen een ‘partikelanalyse’. Zij verwijzen naar andere werkwoorden die wel een direct object hebben, maar die niet passiviseerbaar zijn. Weer anderen, ten slotte, zijn van mening dat passieve varianten van zinnen als (3) wel degelijk mogelijk zijn, op basis waarvan ze concluderen dat op geen postpositie, maar een partikel is. Hoofdstuk 2 concludeert dat eerdere studies tot verschillende analyses zijn gekomen doordat (i) ze niet allemaal gebaseerd zijn op dezelfde tests, (ii) sommige tests tegenstrijdige resultaten opleveren, (iii) grammaticaliteitsoordelen verschillen en (iv) de validiteit van sommige tests ter discussie wordt gesteld.

Dit proefschrift biedt een nieuw perspectief op het oude probleem door middel van een andere methodologie. In plaats van gebruik te maken van syntactische tests veronderstelt het dat de constituentstructuur van constructies als (3) bepaald kan worden op basis van hun semantiek (betekenis). De theoretische achtergrond voor deze veronderstelling is de Cognitieve Grammatica (Langacker 1987, 1991a, 1991b, 2000), waarin constituentstructuur wordt gedefiniëerd als de volgorde waarin eenvoudigere symbolische structuren worden samengenomen (geïntegreerd) om uitgebreidere structuren te vormen. De conclusie van het proefschrift is dat de adpositie in constructies als (3) een partikel is, onderdeel van een scheidbaar samengesteld werkwoord met een speciaal soort direct object. De constructies worden in deze studie dan ook partikelconstructies (ParCs) genoemd.

Een tweede punt waarop dit onderzoek methodologisch afwijkt van eerdere studies is dat het gebaseerd is op authentieke voorbeelden. De meeste zijn afkomstig uit het 38 miljoen woorden corpus van het Instituut voor Nederlandse Lexicologie. Ze zijn aangevuld met voorbeelden uit romans, kranten, conversaties, radio- en televisieprogramma’s en voorbeelden van Internet (verkregen door middel van zoekmachine Google).

Om tot een analyse te komen van de semantiek van ParCs, worden ze in deze studie vergeleken met constructies met preposities zoals (2) (prepositieconstructies, PreCs). De twee typen constructies bieden een uitgelezen kans om de rol van constituentstructuur te bestuderen, omdat ze uit dezelfde elementen bestaan (zie (2) en (3) hierboven): een nominale woordgroep die fungeert als subject (onderwerp), een bewegingswerkwoord, een tweede nominale woordgroep en een adpositie. Het proefschrift veronderstelt dat de semantische verschillen tussen ParCs en PreCs

Constructies, beperkingen en voorstellingswijze 217

gerelateerd kunnen worden aan verschillende manieren waarop deze elementen gecombineerd worden, ofwel aan een verschil in constituentstructuur. Onderdeel van deze veronderstelling is dat de semantiek van een adpositie gelijk blijft in de twee constructies: een betekenisverschil tussen een ParC en een PreC wordt, met andere woorden, niet toegeschreven aan twee verschillende betekenissen van de adpositie.

Hoofdstuk 3 is een kritische bespreking van eerdere studies naar betekenisverschillen tussen ParCs en PreCs. Eerst wordt Kraak en Kloosters (1968) karakterisering in termen van perfectief en imperfectief gepresenteerd en de bezwaren die anderen daartegen hebben geuit. Vervolgens gaat het hoofdstuk in op Helmantel (2002), de tot op heden meest uitgebreide analyse van de semantische verschillen. Haar karakterisering in termen van richting vs. plaats roept echter een aantal vragen op. De vragen betreffen (i) de semantiek van de adposities die in beide constructies voorkomen, (ii) de relatie tussen prePPs en de keuze tussen hebben en zijn en (iii) de semantiek van ParCs. Op deze kwesties wordt achtereenvolgens in hoofdstukken 4, 5 en 6 ingegaan.

Een idee dat in dit proefschrift verkend wordt is dat de semantische structuur van een woord gelijk blijft in verschillende contexten, waarbij de aanname is dat semantische structuur conceptueel is en dat die fungeert als een beperking op het interpretatieproces (Fauconnier 1990, 1994, 1997, Verhagen 1997, 2000). In Hoofdstuk 4 wordt dat uitgewerkt voor een van de adposities die in zowel ParCs als PreCs op kunnen treden, namelijk over. Eerst wordt Cuyckens’ (1991) analyse van over weergegeven, waarin de semantiek van over juist voorgesteld wordt als een netwerk van verschillende, maar verwante betekenissen (senses). Vervolgens wordt beargumenteerd dat veel aspecten van de voorgestelde betekenissen onderdeel zijn van interpretaties van uitgebreidere constructies waarin over voorkomt, maar niet van de semantiek van over. Dit resulteert in een schematischere karakterisering van de semantische structuur van over, met één cruciale toevoeging ten opzichte van eerdere studies naar Nederlands (en Engels) over, namelijk een kracht. Het voorstel is dat over een relatie benoemt tussen een trajector en een landmark (zie voor deze termen Langacker 1987) waarin de trajector gerelateerd is aan de landmark door een mentaal pad dat een oppervlak volgt van de landmark en vanwaar een kracht naar de landmark wijst. In (2), bijvoorbeeld, is een jongen de trajector, die een pad aflegt dat het oppervlak van de landmark de Rijksparallelweg volgt. De kracht van het pad naar de landmark is de zwaartekracht: die trekt de jongen naar de weg. (2) Een jongen [… fietste] op de Rijksparallelweg (mcnov93ove)


In Hoofdstuk 5 wordt onderzocht of er twee typen PreCs

onderscheiden dienen te worden. Helmantel (2002), bijvoorbeeld, onderscheidt twee typen prePPs waarvoor hebben en zijn als diagnostische tests gebruikt kunnen worden: in de sloot in (4), met hebben, heeft een adverbiale functie en duidt de locatie aan waar de beweging plaatsvindt; in de sloot in (5), met zijn, heeft een predicatieve functie en drukt het eindpunt van de beweging uit. (4) dat Jan in de sloot gesprongen heeft (Helmantel 2002: 18) (5) dat Jan in de sloot gesprongen is (Helmantel 2002: 17) Er kleven twee problemen aan dit type analyse. Het eerste is dat de noties ‘locatie’ en ‘eindpunt’ wel van toepassing zijn op prePPs met adposities als in of op (‘punt-adposities’), maar niet op prePPs met adposities als door of over (‘pad-adposities’). Het tweede is dat authentiek taalmateriaal laat zien dat hebben en zijn in feite niet direct correleren met het voorgestelde verschil in prePPs. Vanuit het perspectief van de Cognitieve Grammatica wordt in het hoofdstuk beargumenteerd dat er inderdaad twee typen prePPs zijn (‘modifiers’ en complementen). Hierbij moet opgemerkt worden dat in het geval van pad-adposities de twee typen niet overeenkomen met het onderscheid dat eerder in de literatuur is voorgesteld. Verder correleert de keuze tussen hebben en zijn niet met de twee typen prePPs, maar leggen de hulpwerkwoorden een eigen voorstellingswijze (‘construal’) op: met hebben wordt de beweging voorgesteld als een type handeling, met zijn als een verplaatsing (Geerts et al. 1984, Honselaar 1987, Shannon 1990, Haeseryn et al. 1997). Hoofdstuk 6 vormt het hart van het proefschrift, namelijk de semantische analyse van ParCs, op basis waarvan in het volgende hoofdstuk een analyse van de constituentstructuur gepresenteerd wordt. Het voorstel is dat de ParC een bewegingsproces benoemt waarin een trajector een landmark traverseert zodat resultaat P bereikt wordt: de trajector beweegt van niet P naar helemaal P. ‘P’ in deze beschrijving staat voor de relatie die benoemd wordt door de adpositie. Het hoofdstuk bespreekt de noties traversal (‘overschrijding’), helemaal P en resultaat P stuk voor stuk, waarbij traversal breed gedefinieerd wordt. In (3), bijvoorbeeld, ‘traverseert’ de trajector (een meisje) gedeeltelijk de landmark (de Dokter Larijweg), namelijk vanaf een grens van de landmark, waar ze niet op de landmark is, tot waar ze helemaal op de landmark is, namelijk niet meer ondersteund door iets anders dan de landmark.

Constructies, beperkingen en voorstellingswijze 219

(3) Een […] meisje fietst […] de Dokter Larijweg op (mcaug94ove) ParCs worden in het hoofdstuk voortdurend gecontrasteerd met PreCs, waaruit blijkt dat PreCs niet noodzakelijkerwijs resultatief zijn, zoals (2) laat zien; PreCs vereisen ook niet dat de trajector de landmark traverseert, zoals geïllustreerd kan worden door (6), of dat de trajector uiteindelijk helemaal P is, zie bijvoorbeeld (7). (6) Japie [sprong] op de grond (gp94-2) (7) Het legervoertuig […] reed […] op een mijn (jgdoct93)

In Hoofdstuk 7 wordt het antwoord gegeven op de centrale vraag van deze studie: wat is de constituentstructuur van ParCs? Op basis van drie semantische parallellen wordt geconcludeerd dat een adpositie in een ParC een partikel is: het vormt met het werkwoord een scheidbaar samengesteld werkwoord dat een speciaal soort direct object vereist. De eerste semantische parallel is dat de adpositie van een ParC net als een resultatief partikel van een scheidbaar samengesteld werkwoord het resultaat van de gebeurtenis uitdrukt. De tweede parallel is dat de nominale woordgroep die voorafgaat aan de adpositie in een ParC (geheel of gedeeltelijk) getraverseerd wordt, net als direct objecten van bewegingswerkwoorden in andere talen. De derde parallel is dat ParCs en PreCs een semantisch verschil vertonen dat lijkt op het verschil tussen andere constructies met direct objecten vs. constructies met preposities (ofwel argument/oblique-alternanties, Beavers 2006), namelijk een holistisch/partitief-effect (zie bijvoorbeeld Levin 1993). Deze observaties kunnen verklaard worden als wordt aangenomen dat ParCs en PreCs een andere constituentstructuur hebben.

Hoofdstuk 8 is een evaluatie van de resultaten van de syntactische tests uit Hoofdstuk 2. Nagegaan wordt in hoeverre de resultaten van een test als ondersteuning of probleem voor de gepresenteerde analyse gezien moeten worden. Omdat in sommige gevallen de grammaticaliteitsoordelen uiteenlopen, worden die oordelen eerst getoetst aan authentiek taalmateriaal. Passieve ParCs, bijvoorbeeld, blijken met enige regelmaat voor te komen. Vervolgens wordt beargumenteerd dat een positief grammaticaliteitsoordeel over een constructie niet noodzakelijkerwijs een conclusie rechtvaardigt over de constituentstructuur van een andere constructie. Pas als de twee constructies semantisch overeenkomen, kan gezegd worden dat ze dezelfde constituentstructuur hebben. Er worden suggesties gedaan voor alternatieve analyses van constructies met bijvoorbeeld zijn, r-pronomina en topicalisaties, zodat die niet langer gezien hoeven worden als een probleem voor een partikelanalyse.


In Hoofdstuk 9 worden de resultaten van het proefschrift gepresenteerd en algemenere conclusies getrokken. Deze betreffen het nieuwe (cognitieve-grammatica)perspectief op de constituentstructuur van ParCs, het idee van semantische structuur van een woord of constructie als een beperking op interpretaties, en het belang van de notie ‘voorstellingswijze’ (construal) en van het gebruik van authentiek taalmateriaal. Ten slotte worden vragen voor verder onderzoek geformuleerd, waarvoor dit proefschrift als theoretisch en descriptief vertrekpunt kan dienen.

Curriculum vitae Maaike Beliën was born in Heemskerk, the Netherlands, on August 16, 1972. After finishing secondary school (Bonhoeffer College, Castricum) in 1990, she studied at the University of Exeter (U.K.) for a year. Then she entered the English Language and Literature and General Linguistics programs at the University of Amsterdam, which she completed cum laude in 1996. As part of her studies, she was an exchange graduate student at the UC Berkeley Linguistics Department. From 1998 to 2003 she worked as an ‘Onderzoeker in Opleiding’ at the Vrije Universiteit Amsterdam; in 2001 she spent three months as a visiting scholar at the UC San Diego Linguistics Department. From 2003 to 2008 she was as a lecturer/researcher for the English Language and Culture program at the University of Amsterdam. She is now a lecturer of English as a second language at the Technical University Delft.